Libraries/Microsoft.ML.FastTree.xml
|
<?xml version="1.0"?>
<doc> <assembly> <name>Microsoft.ML.FastTree</name> </assembly> <members> <member name="M:Microsoft.ML.Runtime.FastTree.BoostingFastTreeTrainerBase`2.BsrMaxTreeOutput"> <summary> Retrieves max tree output if best regression step option is active or returns negative value otherwise. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DatasetUtils"> <summary> Loads training/validation/test sets from file </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DatasetUtils.CreateFeatureFromQueryId(Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton)"> <summary> Attempts to create a feature from a ulong array. The intent is that this will handle query ID. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.PerBinStats"> <summary> Holds statistics per bin value for a feature. These are yielded by <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.GetBinStats(System.Int32)"/> to indicate after a <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.Sumup(System.Int32,System.Boolean[],System.Int32,System.Double,System.Double,System.Double[],System.Double[],System.Int32[])"/> call over a subset of the dataset. These statistics are then used in <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner"/> to find splitting on which bin will yield the best least squares solution </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.PerBinStats.SumTargets"> <summary>Sum of all target values in a partition for the bin.</summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.PerBinStats.SumWeights"> <summary>Sum of all target weights in a partition. May be 0 if we are not doing weighted training.</summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.PerBinStats.Count"> <summary>Count of the documents in this partition for the bin.</summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase"> <summary> These objects are stateful, reusable objects that enable the collection of sufficient stats per feature flock, per node or leaf of a tree, to enable it to find the "best" splits. Each instance of this corresponds to a single flock, but multiple of these will be created per flock. Note that feature indices, whenever present, refer to the feature within the particular flock the same as they do with <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase"/>. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.IsSplittable"> <summary> An array as large as there are count of features in the corresponding flock. Used by <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner"/> to indicate whether a particular feature has been judged to be potentially splittable or not. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.Sumup(System.Int32,System.Boolean[],System.Int32,System.Double,System.Double,System.Double[],System.Double[],System.Int32[])"> <summary> Performs the accumulation of sufficient statistics for active features within a flock. </summary> <param name="featureOffset">Offset into <paramref name="active"/> where we should start querying active stats</param> <param name="active">The indicator array of whether features are active or not, logically starting for this flock at <paramref name="featureOffset"/>, where after this </param> <param name="numDocsInLeaf">Minimum documents total in this leaf</param> <param name="sumTargets">The sum of the targets for this leaf</param> <param name="sumWeights">The sum of the weights for this leaf</param> <param name="outputs">The target values, indexed by <paramref name="numDocsInLeaf"/></param> <param name="weights"></param> <param name="docIndices">The first <paramref name="numDocsInLeaf"/> entries indicate the row indices in this leaf, and these row indices are used to </param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.SumupCore(System.Int32,System.Boolean[],System.Int32,System.Double,System.Double,System.Double[],System.Double[],System.Int32[])"> <summary> The core implementation called from <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.Sumup(System.Int32,System.Boolean[],System.Int32,System.Double,System.Double,System.Double[],System.Double[],System.Int32[])"/>. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.Subtract(Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase)"> <summary> Subtracts one sufficient statistics from another. Note that this other sufficient statistics object must be over the same feature flock in order to be meaningful, as well as have undergone <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.Sumup(System.Int32,System.Boolean[],System.Int32,System.Double,System.Double,System.Double[],System.Double[],System.Int32[])"/> under the same set of active features. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.SizeInBytes"> <summary> An approximation of the size in bytes used by this structure. Used for tracking and memory size estimation purposes. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase.CopyFeatureHistogram(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.PerBinStats[]@)"> <summary> Get a fullcopy of histogram for one sub feature. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase`1"> <summary> Type specific implementation of sufficient stats. </summary> <typeparam name="TSuffStats">The type of sufficient stats that we will be able to do "peer" operations against, like subtract. This will always be the derived class itself.</typeparam> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase`1.SubtractCore(`0)"> <summary> Actual core implementation of subtraction. By the time this is called, the base class has ensured that the feature flock of this is the same as the feature flock in the other. </summary> <param name="other">The sufficient statistics we are subtracting</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase"> <summary> A feature flock is a collection of features, grouped together because storing the features and performing the key operations on them in a collection can be done more efficiently than if they were stored as separate features. Since this is a collection of features, feature specific quantities and methods will have a feature index parameter. Note that this index is always, for every flock, from 0 up to but not including <see cref="F:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.Count"/>. Now, in the larger context of a <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset"/> holding many flocks, the individual features might have some sort of "dataset-wide" index, but this is considered the business of the dataset, not the flocks themselves. See <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.MapFeatureToFlockAndSubFeature(System.Int32,System.Int32@,System.Int32@)"/> to see some details of this dataset-wide versus flock-wide feature index. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.Count"> <summary> The number of features contained within this flock. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.Examples"> <summary> The number of training examples represented by the features within this flock. This should be the same for all flocks within a dataset. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.Categorical"> <summary> Flock is a categorical feature. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.SizeInBytes"> <summary> An approximation of the size in bytes used by this structure. Used for estimating memory usage of the tree learner. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.CreateSufficientStats(System.Boolean)"> <summary> A reusable structure for tracking the sufficient statistics for tree learning of the features in this flock. </summary> <param name="hasWeights">Whether structures related to tracking example weights should be allocated</param> <returns>A sufficient statistics object</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GetIndexer(System.Int32)"> <summary> Returns a forward indexer for a single feature. This has a default implementation that relies on <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GetFlockIndexer"/>, but base classes may find it beneficial from a performance perspective to provide their own implementation. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GetFlockIndexer"> <summary> Returns a forward indexer for all features within the flock. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.BinCount(System.Int32)"> <summary> Given a feature index, return the number of distinct bins there are for that feature. This will be the length of <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.BinUpperBounds(System.Int32)"/> for this feature. This is also the upper exclusive limit on the binned value seen for this feature. </summary> <param name="featureIndex">The index of the feature within the flock</param> <returns>The number of distinct bins for that feature within the flock</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.Trust(System.Int32)"> <summary> The multiplier on the gain for any particular feature. This can be used to make features appear more or less attractive. The default value should be considered to be 1. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.BinUpperBounds(System.Int32)"> <summary> An array of increasing values, forming the boundaries of all the binned values. </summary> <param name="featureIndex"></param> <returns>The bin upper bounds for a feature. This array will have the same length as <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.BinCount(System.Int32)"/>.</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GenericIntArrayForwardIndexer"> <summary> If you need to implement <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GetIndexer(System.Int32)"/> you can use <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.GenericIntArrayForwardIndexer"/>. This will be slower than a specialized implementation but is at least a useful shim. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.FlockForwardIndexerBase"> <summary> Interface for objects that can index into a flock, but only with a nondecreasing sequence of row indices from access to access. It is fine for feature indices to appear in any order. A feature group analogy to <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer"/> but for feature flocks instead of <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/> instances. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.FlockForwardIndexerBase.Flock"> <summary> The flock over which this feature flock was built. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase.FlockForwardIndexerBase.Item(System.Int32,System.Int32)"> <summary> Gets the element at the given position. </summary> <param name="featureIndex">The index of the feature within the flock</param> <param name="rowIndex">Index of the row to get, should be non-decreasing from any previous access on this indexer</param> <returns>The value at the index</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1"> <summary> A base class for a feature flock that wraps a single <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/> that contains multiple feature values using a concatentation of the non-zero ranges of each feature, and also in some way that doing a <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Sumup(Microsoft.ML.Runtime.FastTree.Internal.SumupInputData,Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram)"/> will accumulate sufficient statistics correctly for all except the first (zero) bin. </summary> <typeparam name="TIntArray">The type of <see>IntArray</see> this implementation wraps</typeparam> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1.#ctor(`0,System.Int32[],System.Double[][],System.Boolean)"> <summary> Constructor for the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1"/>. </summary> <param name="bins">The binned version of the features, stored collectively in one <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/>, where 0 indicates that all features are in the "cold" bin of 0, and a non-zero value indicates one of the features is "hot," where which feature is hot and what value it has is indicated by <paramref name="hotFeatureStarts"/>. The zero value is "shared" among all features, effectively, and the non-zero values are the result of a shifted concatenation of the range of the non-zero values, for each feature incorporated in the flock. See the example for more concrete information. </param> <param name="hotFeatureStarts">The ranges of values of <paramref name="bins"/> where features start and stop. This is a non-decreasing array of integers. For feature <c>f</c>, the elements at <c>f</c> and <c>f+1</c> indicate the minimum and limit of values in <paramref name="bins"/> that indicate that the corresponding feature is "hot" starting at a value of 1.</param> <param name="binUpperBounds">The bin upper bounds structure</param> <param name="categorical"></param> <example> Imagine we have a six row dataset, with two features, which if stored separately in, say, a <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.SingletonFeatureFlock"/>, would have bin values as follows. <c>f0 = { 0, 1, 0, 0, 2, 0}</c> <c>f1 = { 0, 0, 1, 0, 0, 1}</c> These two are a candidate for a <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.OneHotFeatureFlock"/>, because they never both have a non-zero bin value for any row. Then, in order to represent this in this feature, we would pass in this value for the <paramref name="bins"/>: <c><paramref name="bins"/> = { 0, 1, 3, 0, 2, 3 }</c> and this value for <paramref name="hotFeatureStarts"/>: <c><paramref name="hotFeatureStarts"/> = { 1, 3, 4 }</c> Note that the range of <paramref name="bins"/> is, aside from the zero, the concatenation of the non-zero range of all constituent input features, and where the reconstruction of what feature is which can be reconstructed from <paramref name="hotFeatureStarts"/>, which for each feature specifies the range in <paramref name="bins"/> corresponding to the "logical" bin value for that feature starting from 1. Note that it would also have been legal for <paramref name="hotFeatureStarts"/> to be larger than the actual observed range, e.g., it could have been: <c><paramref name="hotFeatureStarts"/> = { 1, 5, 8}</c> or something. This could happen if binning happened over a different dataset from the data being represented right now, for example, but this is a more complex case. The <paramref name="binUpperBounds"/> would contain the upper bounds for both of these features, which would be arrays large enough so that the maximum value of the logical bin for each feature in the flock could index it. (So in this example, the first bin upper bound would be at least length 3, and the second at least length 2.) The <paramref name="categorical"/> indicates if the flock is a categorical feature. </example> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1.SufficientStats"> <summary> Stores the sufficient statistics for all features within this flock using a single histogram, where the range of what accumulated sufficient statistics are relevant to what feature can be inferred by <see cref="F:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1.HotFeatureStarts"/>. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1.SufficientStats.Hist"> <summary> Stores the sufficient statistics for all features within this flock using a single histogram, where the range of what accumulated sufficient statistics correspond to what actual logical feature depends on <see cref="F:Microsoft.ML.Runtime.FastTree.Internal.SinglePartitionedIntArrayFlockBase`1.HotFeatureStarts"/>. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.NHotFeatureFlock"> <summary> This is a feature flock that misuses a property of <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray.Sumup(Microsoft.ML.Runtime.FastTree.Internal.SumupInputData,Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram)"/> that it works just fine with zero deltas, to "overload" rows into having their targets, thus allowing a single sumup to accumulate multiple "features" on a single row. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.OneHotFeatureFlock"> <summary> A feature flock for a set of features where per example at most one of the features has a non-zero bin. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.OneHotFeatureFlock.Indexer.#ctor(Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer,System.Int32,System.Int32)"> <summary> Instantiates an indexer that translates from the "concatenated" bin space across all features, into the original logical space for each individual feature. </summary> <param name="indexer">The indexer into the "shared" <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/>, that we are translating into the original logical space for this feature, where values in the range of [<paramref name="min"/>,<paramref name="lim"/>) will map from 1 onwards, and all other values will map to 0</param> <param name="min">The minimum value from the indexer that will map to 1</param> <param name="lim">The exclusive upper bound on values from the indexer</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.SingletonFeatureFlock"> <summary> The singleton feature flock is the simplest possible sort of flock, that is, a flock over one feature. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Algorithms.FindFirstGE``1(``0[],``0)"> <summary> Returns the index of the first array position that is larger than or equal to val </summary> <typeparam name="T">an IComparable type</typeparam> <param name="array">a sorted array of values</param> <param name="val">the value to search for</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Algorithms.FindLastLE``1(``0[],``0)"> <summary> Returns the index of the last array position that is less than or equal to val </summary> <typeparam name="T">an IComparable type</typeparam> <param name="array">a sorted array of values</param> <param name="val">the value to search for</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Algorithms.TopK``1(``0[],System.Int32,System.Int32,``0[],System.Int32[])"> <summary> Finds the largest k entries in an array (with offset and length) </summary> <typeparam name="T">An IComparible type</typeparam> <param name="array">The array being searched</param> <param name="offset">An offset into the array</param> <param name="length">The length of the search</param> <param name="topK">The values of the top K</param> <param name="topKPositions">The positions of the top K</param> <returns>The number of entries set in topK and topKPositions (length could be less than K)</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Algorithms.Min``1(``0[],System.Int32@)"> <summary> Fidns the minimum and the argmin in an array of values </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Algorithms.MergeSortedUniqued``1(``0[][])"> <summary> Takes an arbitrary array of sorted uniqued IComparables and returns a sorted uniqued merge </summary> <typeparam name="T">An IComparable </typeparam> <param name="arrays">An array of sorted uniqued arrays</param> <returns>A sorted and uniqued merge</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.BlockingThreadPool"> <summary> This class wraps the standard .NET ThreadPool and adds the following functionality: 1) the user explicitly defines a maximum of concurrently running threads 2) if the maximum is k, and k work items are already running, a call to RunOrBlock will block until a thread is available 3) a work item can be any function with 6 or less arguments 4) a work item knows the index of the thread it is running on - this can be used if the threads share \ common resources </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BlockingThreadPool.Initialize(System.Int32)"> <summary> constructor </summary> <param name="numThreads">the maximal number of concurrent threads</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager"> <summary> This class enables basic buffer pooling. It supports different types of buffers and returns buffers of the requested size or larger. This class was implemented to reduce frequent allocation/deallocation of large buffers which caused fragmentation of the large object heap. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.MinBufferSizeInBytes"> <summary> The minimum size in bytes for a buffer to be stored in the pool This is the minimum size in bytes for an object to be stored in the large object heap </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager._bufferPools"> <summary> A dictionary containing all buffer pool types </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.TakeBuffer``1(System.Int32)"> <summary> Gets a buffer from the pool with at least the same size as passed as input parameter </summary> <typeparam name="T">Pool type</typeparam> <param name="size">Minimum size required</param> <returns>The buffer requested</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.ReturnBuffer``1(``0[]@)"> <summary> Returns a buffer back to the pool. It only keeps buffers bigger than MaxBufferSizeInBytes = 85K bytes </summary> <param name="buffer">The buffer array to add to the pool of buffers</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.ReleaseAllAvailableBuffers(System.Type)"> <summary> Releases all available buffers in a specific pool </summary> <param name="type">Buffer pool type</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.ReleaseAllAvailableBuffers"> <summary> Releases all available buffers in all pools </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.InitializeBufferPool``1"> <summary> Initializes a new buffer pool of a specific type </summary> <typeparam name="T">Type of buffer to initialize</typeparam> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BufferPoolManager.InitializeBufferPool(System.Type)"> <summary> Initializes a new buffer pool of a specific type </summary> <param name="type">Type of buffer to initialize</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LinqExtensions.CumulativeSum``1(System.Collections.Generic.IEnumerable{System.Int32})"> <summary> RunningLength. Converts sequence like 1, 2, 3, 4 to 1, 3, 6, 10 </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.MappedObjectPool`1"> <summary> Implements a paging mechanism on indexed objects. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.MappedObjectPool`1.#ctor(`0[],System.Int32)"> <summary> Initializes a new instance of the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.MappedObjectPool`1"/> class. </summary> <param name="pool">A pool of objects on top of which the paging mechanism is built</param> <param name="maxIndex">The maximal index</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.MappedObjectPool`1.Get(System.Int32,`0@)"> <summary> If the given index maps to a cached object, that object is retrieved and the return value is true. If the index is not cached, an object from the pool is retrieved (possibly paging-out the least-recently used) and the return value is false. </summary> <param name="index">The requested index</param> <param name="obj">The retrieved object</param> <returns>true if the index was found, false if a new object was assigned from the pool</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.MappedObjectPool`1.Reset"> <summary> Resets the MappedObjectPool </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.PseudorandomFunction"> <summary> This class defines a psuedorandom function, mapping a number to a hard to predict but deterministic other number, through some nefarious means. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.StreamExtensions.UndisposableStream"> <summary> A stream class that suppresses the dispose signal </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.StreamExtensions.ReadCompressed(System.IO.Stream)"> <summary> Reads a compressed array of byte from the stream (written by WriteCompressed) </summary> <param name="stream">The stream to read from</param> <returns>The decompressed bytes</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.StreamExtensions.WriteCompressed(System.IO.Stream,System.Byte[],System.Int32,System.Int32)"> <summary> Writes an array of bytes to the stream with compression </summary> <param name="stream">Stream to write to</param> <param name="array">Array to write</param> <param name="offset">The byte offset into the array to write</param> <param name="count">The number of bytes from the array to write</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.StreamExtensions.WriteCompressed(System.IO.Stream,System.Byte[])"> <summary> Writes an array of bytes to the stream with compression </summary> <param name="stream">Stream to write to</param> <param name="array">Array to write</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.ThreadTaskManager.MakeTask(System.Collections.Generic.IEnumerable{System.Action})"> <summary> Makes a new task using the subtasks </summary> <param name="subTasks">subtasks composing the task</param> <returns>An IThreadTask to run the tasks</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.ThreadTaskManager.MakeTask(System.Action{System.Int32},System.Int32)"> <summary> Makes a new task from the supplied action that takes an integer argument, from 0...max </summary> <param name="subTaskAction">Action to run</param> <param name="maxArgument">The max range of the argument</param> <returns>A task that runs the action using each value of the argument from 0...max</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IThreadTask"> <summary> Interface for a decomposable task that runs on many threads </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Timer"> <summary> Static class for timing events. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Timer.TimerState.ToString"> <summary> Gets a string summary of the total times. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Timer.GetTicks(Microsoft.ML.Runtime.FastTree.Internal.TimerEvent)"> <summary> Returns the total number of CPU ticks spent in the specified timer so far. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Timer.Time(Microsoft.ML.Runtime.FastTree.Internal.TimerEvent)"> <summary> Creates a timed event which, when disposed, adds to the total time of that event type. </summary> <param name="e">The type of event</param> <returns>A timed event</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Timer.TimedEvent"> <summary> An object which, when disposed, adds to the total time of that event type. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Timer.GetString"> <summary> Gets a string summary of the total times. </summary> <returns></returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.ToByteArrayExtensions"> <summary> This class contains extension methods that support binary serialization of some base C# types and arrays of these types. SizeInBytes - the number of bytes in the binary representation type.ToByteArray(buffer, ref position) - will write the binary representation of the type to the byte buffer at the given position, and will increment the position to the end of the representation byte[].ToXXX(ref position) - converts the binary representation back into the original type </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.PopulateRawThresholds(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> Converts the bin based thresholds to the raw real-valued thresholds. To be called after training the ensemble. </summary> <param name="dataset">The dataset from which to get the bin upper bounds per feature</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.RemapFeatures(System.Int32[])"> <summary> Remaps the features, to a new feature space. Is called in the event that the features in the training <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset"/> structure are different from the ones in the original pipeline (possibly due to trivialization of input features), and so need to be remapped back to the original space. Note that the tree once modified in this way will no longer have features pointing to the original training <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset"/>, so this should be called only after <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.PopulateRawThresholds(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"/> is called. </summary> <param name="oldToNewFeatures">The mapping from old original features, into the new features</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.ToTreeEnsembleIni(Microsoft.ML.Runtime.FastTree.Internal.FeaturesToContentMap,System.String,System.Boolean,System.Boolean)"> <summary> returns the ensemble in the production TreeEnsemble format </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.GetFeatureContributions(Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,Microsoft.ML.Runtime.Data.BufferBuilder{System.Single}@)"> <summary> Returns a vector of feature contributions for a given example. <paramref name="builder"/> is used as a buffer to accumulate the contributions across trees. If <paramref name="builder"/> is null, it will be created, otherwise it will be reused. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.FeaturesToContentMap"> <summary> A class that given either a <see cref="T:Microsoft.ML.Runtime.Data.RoleMappedSchema"/> provides a mechanism for getting the corresponding input INI content for the features. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeaturesToContentMap.#ctor(Microsoft.ML.Runtime.Data.RoleMappedSchema)"> <summary> Maps input features names to their input INI content based on the metadata of the features column. If the <c>IniContent</c> slotwise string metadata is present, that is used, or else default content is derived from the slot names. </summary> <seealso cref="F:Microsoft.ML.Runtime.Data.MetadataUtils.Kinds.SlotNames"/> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.QuantileRegressionTree.LoadSampledLabels(Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,System.Single[],System.Single[],System.Int32,System.Int32)"> <summary> Loads the sampled labels of this tree to the distribution array for the sparse instance type. By calling for all the trees, the distribution array will have all the samples from all the trees </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.CategoricalSplit"> <summary> Indicates if a node's split feature was categorical. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.CategoricalSplitFeatures"> <summary> Array of categorical values for the categorical feature that might be chosen as a split feature for a node. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.CategoricalSplitFeatureRanges"> <summary> For a given categorical feature that is chosen as a split feature for a node, this array contains it's start and end range in the input feature vector at prediction time. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.TreeType"> <summary> Code to identify the type of tree in binary serialization. These values are persisted, so they should remain consistent for the sake of deserialization backwards compatibility. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.#ctor(System.Int32)"> <summary> constructs a regression tree with an upper bound on depth </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.Create(System.Int32,System.Int32[],System.Double[],System.Single[],System.Single[],System.Int32[],System.Int32[],System.Double[],System.Int32[][],System.Boolean[])"> <summary> Create a Regression Tree object from raw tree contents. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.NumLeaves"> <summary> The current number of leaves in the tree. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.NumNodes"> <summary> The current number of nodes in the tree. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.MaxNumLeaves"> <summary> The maximum number of leaves the internal structure of this tree can support. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.MaxNumNodes"> <summary> The maximum number of nodes this tree can support. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.ScaleOutputsBy(System.Double)"> <summary> Scales all of the output values at the leaves of the tree by a given scalar </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.GetOutput(Microsoft.ML.Runtime.FastTree.Internal.Dataset.RowForwardIndexer.Row)"> <summary> Evaluates the regression tree on a given document. </summary> <param name="featureBins"></param> <returns>the real-valued regression tree output</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.GetOutput(System.Int32[])"> <summary> evaluates the regression tree on a given binnedinstance. </summary> <param name="binnedInstance">A previously binned instance/document</param> <returns>the real-valued regression tree output</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.GetLeafFrom(Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,System.Int32)"> <summary> Returns leaf index the instance falls into, if we start the search from the <paramref name="root"/> node. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.GetLeaf(Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,System.Collections.Generic.List{System.Int32}@)"> <summary> Returns the leaf node for the given feature vector, and populates 'path' with the list of internal nodes in the path from the root to that leaf. If 'path' is null a new list is initialized. All elements in 'path' are cleared before filling in the current path nodes. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.GetOutputs(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> returns the hypothesis output for an entire dataset </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.Split(System.Int32,System.Int32,System.Int32[],System.Int32[],System.Boolean,System.UInt32,System.Double,System.Double,System.Double,System.Double)"> <summary> Turns a leaf of the tree into an interior node with two leaf-children. </summary> <param name="leaf">The index of the leaf to split.</param> <param name="feature">The index of the feature used to split this leaf (as it indexes the array of DerivedFeature instances passed to the to tree ensemble format).</param> <param name="categoricalSplitFeatures">Thresholds for categorical split.</param> <param name="categoricalSplitRange"></param> <param name="categoricalSplit"></param> <param name="threshold">The </param> <param name="lteValue">The value of the leaf on the LTE side.</param> <param name="gtValue">The value of the leaf on the GT side.</param> <param name="gain">The splitgain of this split. This does not affect the logic of the tree evaluation.</param> <param name="gainPValue">The p-value associated with this split, indicating confidence that this is a better than random split. This does not affect the logic of the tree evaluation.</param> <returns>Returns the node index</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.ToTreeEnsembleFormat(System.Text.StringBuilder,System.Text.StringBuilder,Microsoft.ML.Runtime.FastTree.Internal.FeaturesToContentMap,System.Int32@,System.Collections.Generic.Dictionary{System.Int32,System.Int32})"> <summary> Returns a representation of the tree in the production decision tree format (SHAREDDYNAMICRANKROOT\TreeEnsembleRanker\Tree.h). The intent is that this </summary> <param name="sbEvaluator">Append the new evaluator to this stringbuilder.</param> <param name="sbInput">Append any hitherto unused [Input:#] sections to this stringbuilder.</param> <param name="featureContents">The feature to content map.</param> <param name="evaluatorCounter">A running count of evaluators. When this method returns it should have one more entry.</param> <param name="featureToId">A map of feature index (in the features array) to the ID as it will be written in the file. This instance should be used for all </param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.OptimizePathToLeaf(System.Int32)"> <summary> Sets the path to a leaf to be indexed by 0,1,2,3,... and sets the leaf index to 0 </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTree.SwapNodePositions(System.Int32,System.Int32)"> <summary> swaps the positions of two nodes in the tree, without any functional change to the tree </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.TrivialGradientWrapper"> <summary> Trivial weights wrapper. Creates proxy class holding the targets. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.BestStepRegressionGradientWrapper"> <summary> Provides weights used when best regression step option is on. </summary> Second-derivatives used as weights in a leaf when one makes Newton-Raphson step (taken in account when regression tree is trained). </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.QueryWeightsGradientWrapper"> <summary> Wraps targets with query weights. Regression tree is built for weighted data, and weights are used for mean calculation at Newton-Raphson step. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.QueryWeightsBestResressionStepGradientWrapper"> <summary> Wraps targets whan both query weights and best step regression options are active. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator._secondLevelcomparers"> <summary> Contains the instances for a second Level comparer, which gets applied after the initial rank based ordering has happened. The array stores one second level comparer per thread. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.#ctor(System.Int32,System.String,System.Int32)"> <summary> Constructs a DCG calculator </summary> <param name="maxDocsPerQuery">the maximum number of documents per query</param> <param name="sortingAlgorithm">a string describing the sorting algorithm to use</param> <param name="topNDocsForIdealDcg">specifies the ideal DCG@ computation.</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.MaxDcg(System.Int16[],System.Int32[],System.Int32,System.Int32[][])"> <summary> Calculates the natural-based max DCG at a given truncation </summary> <param name="labels">vector of labels</param> <param name="boundaries">vector of query boundaries</param> <param name="trunc">truncation to use</param> <param name="labelCounts"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.MaxDcgQuery(System.Int16[],System.Int32,System.Int32,System.Int32,System.Int32[])"> <summary> Calculates the natural-based max DCG at a given truncation for a query </summary> <param name="labels">vector of labels</param> <param name="begin">Index of the first document</param> <param name="labelCounts"></param> <param name="trunc">truncation to use</param> <param name="numDocuments"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.Ndcg3(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int16[],System.Double[])"> <summary> Efficient computation of average NDCG@3 for the entire dataset Note that it is virtual and MPI provides faster implementations for MPI </summary> <param name="dataset">the dataset</param> <param name="scores">vector of scores</param> <param name="labels"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.DCG3(System.Double[],System.Int16[],System.Int32,System.Int32)"> <summary> Efficient computation of natural-based pessimistic DCG@3 for a given query </summary> <param name="scores">vector of scores</param> <param name="labels">vector of labels</param> <param name="begin">index of first document in query</param> <param name="end">index of first document in next query</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.Ndcg1(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int16[],System.Double[])"> <summary> Efficient computation of average NDCG@1 for the entire dataset Note that it is virtual and MPI provides faster implemetations for MPI </summary> <param name="dataset">the dataset</param> <param name="labels"></param> <param name="scores">the vector of score from previous rounds</param> <returns>average NDCG@1 for an entire dataset</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.DCG1(System.Double[],System.Int16[],System.Int32,System.Int32)"> <summary> Calculates the natural-based pessimistic DCG@1 of scores(query) </summary> <param name="scores">vector of scores</param> <param name="labels">vector of labels</param> <param name="begin">index of first document in query</param> <param name="end">index of first document in next query</param> <returns>DCG@1</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.NdcgRangeFromScores(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int16[],System.Double[])"> <summary> calculates the average NDCG given the scores array For performance reason it duplicates some </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.SortRankingResults(System.Int16[],System.Int32,System.Int32,System.Int32,System.Int32,System.Int32[],System.Double[])"> <summary> Orders the queries based on the given comparer. </summary> <param name="labels">The label for all query URL pairs</param> <param name="threadIndex">Specifies the thread which is executing this code</param> <param name="scoreBegin">position of the first query-URL pair to sort in the score array</param> <param name="labelBegin">position of the first query-URL pair to sort in the label array</param> <param name="count">number of query-URL pairs</param> <param name="permutation">resulting query order array</param> <param name="scores">The scores for all query URL pairs</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DcgCalculator.OrderingFromScores(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Double[])"> <summary> Calculates the order of documents. This returns an array with as many elements as there are documents, where the subarray in a query's boundary will contain elements from 0 up to but not including the number of documents in the query. The first value in this subarray will contain the index of the document in the subarray at top position (highest ranked), and the last value the index of the document with bottom position (lowest ranked). </summary> <param name="dataset">The dataset over which to calculate the DCG.</param> <param name="scores">The scores for all documents within the dataset.</param> <returns></returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DescendingStablePessimisticPermutationComparer"> <summary> Compares two integers that are indices into a vector of doubles. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DescendingStablePermutationComparer"> <summary> Compares two integers that are indices into a vector of doubles. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer"> <summary> Implements an HRS based comparer to sort the ranking results for the first N results. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer.#ctor(System.Int32)"> <summary> Creates an instance of the DescendingStableIdealComparer for the TOP N query/URL pairs </summary> <param name="comparefirstN">Specifies the TOP N query/URL pairs which should be used for sorting</param> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer.CompareFirstN"> <summary> Specifies the TOP N query/URL pairs which should be used for sorting </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer.Labels"> <summary> The HRS labels for all query/URL pairs </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer.LabelsOffset"> <summary> The position inside the Labels where the this query starts for which the URLs should be reshuffled. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DescendingStableIdealComparer.System#Collections#Generic#IComparer{System#Int32}#Compare(System.Int32,System.Int32)"> <summary> Compare two HRS ratings for query/URL pairs </summary> <param name="x">position for query/URL pair 1</param> <param name="y">position for query/URL pair 2</param> <returns></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.#ctor(System.Int32,System.Int32)"> <summary> Constructor </summary> <param name="numDocuments">number of document</param> <param name="maxLeaves">number of leaves</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.#ctor(Microsoft.ML.Runtime.FastTree.Internal.RegressionTree,Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> Constructs partitioning object based on the documents and RegressionTree splits NOTE: It has been optimized for speed and multiprocs with 10x gain on naive LINQ implementation </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.NumDocs"> <summary> Returns the total number of documents handled by the partitioning </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.Initialize"> <summary> Resets the data structure, as if it was newly created </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.GetDistribution(System.Double[],System.Double[],System.Int32,System.Random,System.Int32,System.Double[]@)"> <summary> Does sampling with replacement on each leaf node and returns leaf count times of sample labels. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.Split(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer,System.UInt32,System.Int32)"> <summary> Splits the documents of a specified leaf to its two children based on a feature and a threshold value </summary> <param name="leaf">the leaf being split</param> <param name="indexer"></param> <param name="threshold">the threshold</param> <param name="gtChildIndex">Index of child node that contains documents whose split feature value is greater than the split threshold</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.Split(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.IntArray,System.Collections.Generic.HashSet{System.Int32},System.Int32)"> <summary> Splits the documents of a specified leaf to its two children based on a feature and a threshold value </summary> <param name="leaf">the leaf being split</param> <param name="bins">Split feature flock's bin</param> <param name="categoricalIndices">Catgeorical feature indices</param> <param name="gtChildIndex">Index of child node that contains documents whose split feature value is greater than the split threshold</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.GetLeafDocumentPartitions(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer,System.UInt32,System.Int32[]@,System.Int32@,System.Int32[]@,System.Int32@)"> <summary> Get the document partitions of a specified leaf if it is split based on a feature and a threshold value. </summary> <param name="leaf">the leaf being split</param> <param name="indexer">the indexer to access the feature value</param> <param name="threshold">the threshold</param> <param name="leftDocuments">[out] the left documents split from the leaf</param> <param name="leftDocumentSize">[out] the size of left documents</param> <param name="rightDocuments">[out] the right documents split from the leaf</param> <param name="rightDocumentSize">[out] the size of right documents</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.DocumentsInLeaf(System.Int32)"> <summary> Returns an enumeration of the document indices associated with a specified leaf, in ascending order </summary> <param name="leaf">the leaf index</param> <returns>the enumeration</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.NumDocsInLeaf(System.Int32)"> <summary> How many documents are associated with a specified leaf </summary> <param name="leaf">the leaf</param> <returns>the number of documents</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.Mean(System.Double[],System.Int32,System.Boolean)"> <summary> Calculates the mean of a double array only on the elements that correspond to a specified leaf in the tree </summary> <param name="array">the double array</param> <param name="leaf">the leaf index</param> <param name="filterZeros"></param> <returns>the mean</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning.Mean(System.Double[],System.Double[],System.Int32,System.Boolean)"> <summary> Calculates the weighted mean of a double array only on the elements that correspond to a specified leaf in the tree </summary> <param name="array">the double array</param> <param name="sampleWeights">Weights of array elements</param> <param name="leaf">the leaf index</param> <param name="filterZeros"></param> <returns>the mean</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.LassoBasedEnsembleCompressor"> <summary> This implementation is based on: Friedman, J., Hastie, T. and Tibshirani, R. (2008) Regularization Paths for Generalized Linear Models via Coordinate Descent. http://www-stat.stanford.edu/~hastie/Papers/glmnet.pdf </summary> <remarks>Author was Yasser Ganjisaffar during his internship.</remarks> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.GradientDescent._numberOfDroppedTrees"> number of trees dropped in this iteration </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IGradientAdjuster"> <summary> Interface for wrapping with weights of gradient target values </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IGradientAdjuster.AdjustTargetAndSetWeights(System.Double[],Microsoft.ML.Runtime.FastTree.Internal.ObjectiveFunctionBase,System.Double[]@)"> <summary> Create wrapping of gradient target values </summary> <param name="gradient"></param> <param name="objFunction">Objective functions can be used for constructing weights</param> <param name="targetWeights"></param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.RandomForestOptimizer"> <summary> This is dummy optimizer. As Random forest does not have any boosting based optimization, this is place holder to be consistent with other fast tree based applications </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.TestResult.RawValue"> <summary> Raw value used for calculating final test result value. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.TestResult.Factor"> <summary> The factor used for calculating final test result value. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.TestResult.Operator"> <summary> The operator used for calculating final test result value. Final value = Operator(RawValue, Factor) </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.TestResult.LowerIsBetter"> <summary> Indicates that the lower value of this metric is better This is used for early stopping (with TestHistory and TestWindowWithTolerance) </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.RegressionTest.#ctor(Microsoft.ML.Runtime.FastTree.Internal.ScoreTracker,System.Nullable{System.Int32})"> <param name="scoreTracker"></param> <param name="resultType">1: L1, 2: L2. Otherwise, return all.</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner"> <summary> Trains regression trees. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.#ctor(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int32,System.Int32,System.Double,System.Double,System.Double,System.Double,System.Int32,System.Int32,System.Double,System.Boolean,System.Boolean,System.Double,System.Int32,System.Int32,System.Double,Microsoft.ML.Runtime.FastTree.IParallelTraining,System.Double,Microsoft.ML.Runtime.FastTree.Bundle,System.Int32,System.Double)"> <summary> Creates a new LeastSquaresRegressionTreeLearner </summary> <param name="trainData">Data to train from</param> <param name="numLeaves">Maximum leaves in tree</param> <param name="minDocsInLeaf">Minimum allowable documents in leaf</param> <param name="entropyCoefficient">Add the information gain of a split to the gain times this value. Practically, this will favor more balanced splits</param> <param name="featureFirstUsePenalty">Features never used before effectively have this amount subtracted from their gain</param> <param name="featureReusePenalty">Features used before effectively have this amount subtracted from their gain</param> <param name="softmaxTemperature">Regularization parameter, where we become increasingly likely to select a non-optimal split feature the higher the temperature is</param> <param name="histogramPoolSize">Number of feature histograms to cache</param> <param name="randomSeed">The seed to use for sampling</param> <param name="splitFraction"></param> <param name="filterZeros">Whether we should ignore zero lambdas for the purpose of tree learning (generally a bad idea except for when zero indicates that the document should be ignored)</param> <param name="allowEmptyTrees">If false, failure to split the root will result in error, or if true will result in null being returned when we try to fit a tree</param> <param name="gainConfidenceLevel">Only consider a gain if its likelihood versus a random choice gain is above a certain value (so 0.95 would mean restricting to gains that have less than a 0.05 change of being generated randomly through choice of a random split).</param> <param name="maxCategoricalGroupsPerNode">Maximum categorical split points to consider when splitting on a categorical feature.</param> <param name="maxCategoricalSplitPointPerNode"></param> <param name="bsrMaxTreeOutput">-1 if best step ranking is to be disabled, otherwise it is interpreted as being similar to the maximum output for a leaf</param> <param name="parallelTraining"></param> <param name="minDocsPercentageForCategoricalSplit"></param> <param name="bundling"></param> <param name="minDocsForCategoricalSplit"></param> <param name="bias"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FitTargets(Microsoft.ML.Runtime.IChannel,System.Boolean[],System.Double[])"> <summary> Learns a new tree for the current outputs </summary> <returns>A regression tree</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.Initialize(System.Boolean[])"> <summary> Clears data structures </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FindBestSplitOfRoot(System.Double[])"> <summary> finds best feature/threshold split of root node. Fills in BestSplitInfoPerLeaf[0] </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FindBestSplitOfSiblings(System.Int32,System.Int32,Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning,System.Double[])"> <summary> Finds best feature/threshold split of <paramref name="lteChild"/> and <paramref name="gtChild"/>, and fills in the corresponding elements of <see cref="F:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.BestSplitInfoPerLeaf"/>. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FindAndSetBestFeatureForLeaf(Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates)"> <summary> After the gain for each feature has been computed, this function chooses the gain maximizing feature and sets its info in the right places This method is overriden in MPI version of the code </summary> <param name="leafSplitCandidates">the FindBestThesholdleafSplitCandidates data structure that contains the best split information</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FindBestThresholdForFlockThreadWorker(System.Int32)"> <summary> The multithreading entry-point: finds the best threshold for a given flock at a given leaf. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.GetActiveFeatures(System.Int32,System.Int32)"> <summary> Returns the set of features that are active within a particular range. </summary> <param name="min">The inclusive lower bound of the feature indices</param> <param name="lim">The exclusive upper bound of the feature indices</param> <returns>The feature indices within the range that are active</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.GetLeafSplitGain(System.Int32,System.Double,System.Double)"> <summary> Returns the split gain for a particular leaf. Used on two leaves to calculate the squared error gain for a particular leaf. </summary> <param name="count">Number of documents in this leaf</param> <param name="sumTargets">Sum of the target values for this leaf</param> <param name="sumWeights">Sum of the weights for this leaf, not meaningful if <see cref="P:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.HasWeights"/> is <c>false</c></param> <returns>The gain in least squared error</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.CalculateSplittedLeafOutput(System.Int32,System.Double,System.Double)"> <summary> Calculates the output value for a leaf after splitting. </summary> <param name="count">Number of documents in this leaf</param> <param name="sumTargets">Sum of the target values for this leaf</param> <param name="sumWeights">Sum of the weights for this leaf, not meaningful if <see cref="P:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.HasWeights"/> is <c>false</c></param> <returns>The output value for a leaf</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.FindBestThresholdFromHistogram(Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase,Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates,System.Int32)"> <summary> Finds the best threshold to split on, and sets the appropriate values in the LeafSplitCandidates data structure. </summary> <param name="histogram">The sufficient stats accumulated for the flock</param> <param name="leafSplitCandidates">The LeafSplitCandidates data structure</param> <param name="flock">The index of the flock containing this feature</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates"> <summary> Contains the memory data structures required for finding the best threshold for a given feature at a given leaf. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates.Initialize(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning,System.Double[],System.Double[],System.Boolean)"> <summary> Initializes the object for a specific leaf, with a certain subset of documents. </summary> <param name="leafIndex">The leaf index</param> <param name="partitioning">The partitioning object that knows which documents have reached that leaf</param> <param name="targets">The array of targets, which the regression tree is trying to fit</param> <param name="weights">The array of weights for the document</param> <param name="filterZeros">Whether filtering of zero gradients was turned on or not</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates.Initialize(System.Double[],System.Double[],System.Boolean)"> <summary> Initializes the object for computing the root node split </summary> <param name="targets">the array of targets, which the regression tree is trying to fit</param> <param name="weights"></param> <param name="filterZeros"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates.Initialize"> <summary> Initializes the object to do nothing </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.SplitInfo"> <summary> A struct to store information about each leaf for splitting </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TreeLearner.GetSizeOfReservedMemory"> <summary> Get size of reserved memory for the tree learner. The default implementation returns 0 directly, and the subclasses can return different value if it reserves memory for training. </summary> <returns>size of reserved memory</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.TreeLearnerException"> <summary> An exception class for an error which occurs in the midst of learning a tree. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.WinLossCalculator.#ctor(System.Int32,System.String)"> <summary> Constructs a WinLoss calculator </summary> <param name="maxDocsPerQuery">the maximum number of documents per query</param> <param name="sortingAlgorithm">a string describing the sorting algorithm to use</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.WinLossCalculator.WinLossRangeFromScores(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int16[],System.Double[])"> <summary> calculates the average WinLoss given the scores array For performance reason it duplicates some </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset"> <summary> A dataset of features. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.#ctor(Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton,Microsoft.ML.Runtime.FastTree.Internal.FeatureFlockBase[])"> <summary> Initializes a new instance of the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset"/> class. </summary> <param name="datasetSkeleton">The dataset skeleton corresponding to the features</param> <param name="flocks">An array of feature flocks</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.MapFeatureToFlockAndSubFeature(System.Int32,System.Int32@,System.Int32@)"> <summary> Maps a global feature index, to the index of the particular flock, as well as the index of the subfeature within that flock. </summary> <param name="feature">The index of the feature at the dataset level</param> <param name="flock">The index of the flock containing this feature</param> <param name="subfeature">The index of the feature within the flock</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.FlockToFirstFeature(System.Int32)"> <summary> Given a flock index, returns the index of the first feature in this flock. </summary> <param name="flock">Index of the flock</param> <returns>The index of the first feature that belongs to this flock</returns> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.Skeleton"> <summary> Gets the dataset skeleton. </summary> <value>The skeleton.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.Ratings"> <summary> Gets the labels. </summary> <value>The labels.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.Boundaries"> <summary> Gets the boundaries. </summary> <value>The boundaries.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.QueryIds"> <summary> Gets the query ids. </summary> <value>The query ids.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DocIds"> <summary> Gets the doc ids. </summary> <value>The doc ids.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.MaxDcg"> <summary> Gets the max DCG. </summary> <value>The max DCG.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.MaxDocsPerQuery"> <summary> Gets the max number of docs per any query. </summary> <value>The max number of docs per any query.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.NumDocs"> <summary> Gets the number of docs in the entire dataset. </summary> <value>The number of docs in the entire dataset.</value> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.NumDocsInQuery(System.Int32)"> <summary> Nums the docs in a given query. </summary> <param name="queryIndex">Index of the query.</param> <returns>the number of docs in the query</returns> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.NumQueries"> <summary> Gets the number of queries in the dataset. </summary> <value>The number of queries in the dataset.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DocToQuery"> <summary> Returns the document to query </summary> <returns>The associated document</returns> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.SampleWeights"> <summary> Returns the query weights object in underlying dataset skeleton </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.Flocks"> <summary> Gets the array of features. </summary> <value>The array of features.</value> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.NumFlocks"> <summary> The number of feature flocks. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.NumFeatures"> <summary> The number of features. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.Split(System.Double[],System.Int32,System.Boolean)"> <summary> Split a dataset by queries into disjoint parts </summary> <param name="fraction">an array of the fractional size of each part, must sum to 1.0</param> <param name="randomSeed">a seed that deterministically defines the split</param> <param name="destroyThisDataset">do you want the features of this dataset to be destroyed on-the-fly as the new datasets are created</param> <returns></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.GetSubDataset(System.Int32[],System.Boolean)"> <summary> Creates a new Dataset, which includes a subset of the docs in this Dataset. </summary> <param name="docIndices">A sorted array of doc indices</param> <param name="destroyThisDataset">Determines if this Dataset is deleted on the fly as the new one is created (this reduces peak memory)</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.GetFeatureBinRowwiseIndexer(System.Boolean[])"> <summary> Returns a row-wise forward indexer across multiple features in the dataset. </summary> <param name="activeFeatures">Boolean array indicating active features, or null to indicate all features should be used</param> <returns>Row forward indexer</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton"> <summary> A class that contains all of the feature-independent data of the dataset </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.#ctor(System.Int16[],System.Int32[],System.UInt64[],System.UInt64[],System.Double[])"> <summary> Initializes a new instance of the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton"/> class. </summary> <param name="ratings"></param> <param name="boundaries">The boundaries.</param> <param name="queryIds">The query ids.</param> <param name="docIds">The doc ids.</param> <param name="actualTargets"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.#ctor(System.Int16[],System.Int32[],System.UInt64[],System.UInt64[],System.Double[][],System.Double[])"> <summary> Initializes a new instance of the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton"/> class. </summary> <param name="ratings">The ratings.</param> <param name="boundaries">The boundaries.</param> <param name="queryIds">The query ids.</param> <param name="docIds">The doc ids.</param> <param name="maxDcg">The vector of maxDCG.</param> <param name="actualTargets"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.CheckConsistency"> <summary> Checks the consistency of the DatasetSkeleton </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.Concat(Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton[])"> <summary> Takes an array of DatasetSkeleton objects and concatenates them into one big DatasetSkeleton </summary> <param name="parts">An array of DatasetSkeletons</param> <returns>A concatenated DatasetSkeleton</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.MaxDcgRange(System.Int16[],System.Int32[],System.Int32)"> <summary> Calculates natural-based max DCG at all truncations from 1 to trunc </summary> <param name="labels">vector of labels</param> <param name="boundaries">vector of query boundaries</param> <param name="trunc">max truncation</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.SetConcatenatedAuxiliaryData(Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton[],Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton)"> <summary> Given the auxiliary data in a bunch of parts, set the concatenated dataset appropriately. </summary> <param name="parts">The individual parts of the dataset</param> <param name="concat">The concatenated version of this dataset</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.SetData(System.String,System.Array,System.Boolean)"> <summary> Sets some named query or document level auxiliary data. </summary> <param name="name">The name of the parameter</param> <param name="array"></param> <param name="queryLevel"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.DatasetSkeleton.GetData``1(System.String)"> <summary> Retrieves some auxiliary data previously set to this skeleton. </summary> <typeparam name="T">The type of the array, which should match the type passed in</typeparam> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dataset.RowForwardIndexer"> <summary> Structure allowing forward indexing by row, across multiple features in the dataset. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Dataset.RowForwardIndexer.Row.Item(System.Int32)"> <summary> Indexes the value of a feature for this row. </summary> <param name="featureIndex">The feature index</param> <returns>The binned valued of a feature for this row</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dataset.RowForwardIndexer.#ctor(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Boolean[])"> <summary> Constructor. </summary> <param name="dataset">The dataset to create the indexer over</param> <param name="active">Either null to indicate all columns should be active, or a boolean array of length equal to the number of features that should be active</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DenseIntArray"> <summary> Abstract class implementing some common functions of the dense int array types. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.DenseIntArray.Item(System.Int32)"> <summary> Gets or sets the value at this index. Value must be in legal range 0...((2^<see cref="P:Microsoft.ML.Runtime.FastTree.Internal.IntArray.BitsPerItem"/>)-1). </summary> <param name="index">Index of value to get or set</param> <returns>The value at this index</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DenseIntArray.Clone(System.Int32[])"> <summary> Clone an IntArray containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned IntArray </param> <returns> The cloned IntArray </returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense0BitIntArray"> <summary> A "null" feature representing only zeros. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense0BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense0BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense10BitIntArray"> <summary> A class to represent features using 10 bits. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense10BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense10BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense8BitIntArray"> <summary> A class to represent features using 8 bits </summary> <remarks>Represents values -1...(2^s-2) 0-bit array only represents the value -1</remarks> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense8BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense8BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense4BitIntArray"> <summary> A class to represent features using 4 bits. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.Dense4BitIntArray._data"> <summary> For a given byte, the high 4 bits is the first value, the low 4 bits is the next value. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense4BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense4BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense16BitIntArray"> <summary> A class to represent features using 16 bits. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense16BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense16BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Dense32BitIntArray"> <summary> A class to represent features using 32 bits. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense32BitIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Dense32BitIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Feature"> <summary> Represents a binned feature </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.Feature.FeatureType"> <summary> The type of the feature. This is serialized as part of the Bing extraction BIN file, so it should remain binary compatible from version to version. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Feature.Length"> <summary> Returns the number of entires (documents) in the feature </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Feature.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.Feature.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.Feature.Bins"> <summary> Gets the compactIntArray of bin values. </summary> <value>The bin values.</value> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.#ctor(Microsoft.ML.Runtime.FastTree.Internal.IntArray,System.UInt32[],System.String)"> <summary> Initializes a new instance of the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.Feature"/> class. </summary> <param name="bins">The bins.</param> <param name="valueMap"></param> <param name="name">The name.</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.#ctor(System.String,System.Int32)"> <summary> Constructs an empty (all zero) feature </summary> <param name="name"></param> <param name="length"></param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.ValueMap"> <summary> Gets the value that represents each bin </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.Clone(System.Int32[])"> <summary> Clone a TSVFeature containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned TSVFeature </param> <returns> The cloned TSVFeature </returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.TsvFeature.Concat(Microsoft.ML.Runtime.FastTree.Internal.TsvFeature[])"> <summary> Concatenates an array of features into one long feature </summary> <param name="parts">An array of features</param> <returns>A concatenated feature</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram"> <summary> Class to represent statistics of the feature used by LeastSquaresRegressionTreeLearner </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram.#ctor(Microsoft.ML.Runtime.FastTree.Internal.IntArray,System.Int32,System.Boolean)"> <summary> Make a new FeatureHistogram </summary> <param name="bins">The bins we will be calculating sumups over</param> <param name="numBins">The number of bins, should be at least as large as the number of bins</param> <param name="useWeights">Allocates weights array when true</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram.EstimateMemoryUsedForFeatureHistogram(System.Int32,System.Boolean)"> <summary> This function returns the estimated memory used for a FeatureHistogram object according to given number of bins. </summary> <param name="numBins">number of bins</param> <param name="hasWeights">weights array is counted when true</param> <returns>estimated size of memory used for a feature histogram object</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram.Subtract(Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram)"> <summary> Subtract from myself the counts of the child histogram </summary> <param name="child">Another histogram to subtract</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"> <summary> An object representing an array of integers </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Length"> <summary> The virtual length of the array </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary representation is written</param> <param name="position">the position in the byte array</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.NumBitsNeeded(System.Int32)"> <summary> Number of bytes needed to store this number of values </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.New(System.Byte[],System.Int32@)"> <summary> Creates a new int array given a byte representation </summary> <param name="buffer">the byte array representation of the dense array. The buffer can be larger than needed since the caller might be re-using buffers from a pool</param> <param name="position">the position in the byte array</param> <returns>the int array object</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Clone(Microsoft.ML.Runtime.FastTree.Internal.IntArrayBits,Microsoft.ML.Runtime.FastTree.Internal.IntArrayType)"> <summary> Clones the contents of this IntArray into an new IntArray </summary> <param name="bitsPerItem">The number of bits per item in the created IntArray</param> <param name="type">The type of the new IntArray</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Clone(System.Int32[])"> <summary> Clone an IntArray containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned IntArray </param> <returns> The cloned IntArray </returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.GetIndexer"> <summary> Gets an indexer into the array </summary> <returns>An indexer into the array</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Compress(System.UInt32[])"> <summary> Finds the most space efficient representation of the feature (with slight slack cut for dense features). The behavior of this method depends upon the static value <see cref="F:Microsoft.ML.Runtime.FastTree.Internal.IntArray.CompatibilityLevel"/>. </summary> <param name="workarray">Should be non-null if you want it to consider segment arrays.</param> <returns>Returns a more space efficient version of the array, or the item itself if that is impossible, somehow.</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer"> <summary> Interface for objects that can index into an <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/>, but only with a non-decreasing sequence of indices. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.Internal.IIntArrayForwardIndexer.Item(System.Int32)"> <summary> Gets the element at the given index. </summary> <param name="index">Index to get</param> <returns>The value at the index</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaRepeatIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaRepeatIntArray.Clone(System.Int32[])"> <summary> Clone an IntArray containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned IntArray </param> <returns> The cloned IntArray </returns> </member> <member name="F:Microsoft.ML.Runtime.FastTree.Internal.SegmentIntArray.TransitionCost"> <summary> The cost of a transition between segments in bits. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SegmentIntArray.StatsOfBestEncoding(System.UInt32[],System.Int32,System.Boolean,System.Int64@,System.Int32@,System.UInt32@)"> <summary> Finds the bits necessary for the optimal variable bit encoding of this array. If we are also finding the actual optimal path, it can only work This is a considerably less efficienct managed analogue to the C_SegmentFindOptimalPath and C_SegmentFindOptimalCost functions. It is used by the class only when not using the unmanaged library. </summary> <param name="ivalues">The values for which we should find the optimal cost. If findPath is active, the most significant 5 bits will be used to store the bitness with which this path should be chosen.</param> <param name="bitsForMaxItem">This should be the maximum number of bits necessary to encode the largest item in that array, or a higher value. Owing to the nature of the values as 32 bit quantities this value should be in the range [0,32], or 21 if we are finding the cannot exceed 31.</param> <param name="findPath">Whether we should find the best path, by also storing the optimal path in the most 5 significant digits.</param> <param name="bits">The number of bits necessary for the optimal encoding.</param> <param name="transitions">The number of transitions necessary in the optimal encoding (only if findPath is true).</param> <param name="max">The maximum element in the ivalues array.</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.SegmentIntArray.Clone(System.Int32[])"> <summary> Clone an IntArray containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned IntArray </param> <returns> The cloned IntArray </returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray"> <summary> This <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.IntArray"/> implementation represents a sequence of values using parallel arrays of both values, as well as deltas indicating the number of values to the next explicit value. Values "between" these deltas are implicitly zero. Note that it is possible to misuse the deltas by making some of them themselves 0, allowing us to represent multiple values per row. In this case, <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.GetEnumerator"/> and <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.GetIndexer"/> will not have sensible values, but <see cref="M:Microsoft.ML.Runtime.FastTree.Internal.IntArray.Sumup(Microsoft.ML.Runtime.FastTree.Internal.SumupInputData,Microsoft.ML.Runtime.FastTree.Internal.FeatureHistogram)"/> will work and accumulate the same target into multiple bins. This trick should be used with caution, and is useful for the <see cref="T:Microsoft.ML.Runtime.FastTree.Internal.NHotFeatureFlock"/>. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray.#ctor(System.Int32,Microsoft.ML.Runtime.FastTree.Internal.IntArrayBits,System.Collections.Generic.IEnumerable{System.Collections.Generic.KeyValuePair{System.Int32,System.Int32}})"> <summary> Construct a sparse int array from index, value pairs. </summary> <param name="length">The total length of the constructed array.</param> <param name="bitsPerItem">The number of bits required to store the values.</param> <param name="nonZeroValues">An ordered enumerable of (index,value) pairs. Each index should be strictly increasing as the iterable proceeds.</param> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray.Clone(System.Int32[])"> <summary> Clone an IntArray containing only the items indexed by <paramref name="itemIndices"/> </summary> <param name="itemIndices"> item indices will be contained in the cloned IntArray </param> <returns> The cloned IntArray </returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray.SizeInBytes"> <summary> Returns the number of bytes written by the member ToByteArray() </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.DeltaSparseIntArray.ToByteArray(System.Byte[],System.Int32@)"> <summary> Writes a binary representation of this class to a byte buffer, at a given position. The position is incremented to the end of the representation </summary> <param name="buffer">a byte array where the binary represenaion is written</param> <param name="position">the position in the byte array</param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.BinFinder"> <summary> A class that bins vectors of doubles into a specified number of equal mass bins. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BinFinder.FindDistinctCounts(Microsoft.ML.Runtime.Data.VBuffer{System.Double}@,System.Double[],System.Int32[])"> <summary> Finds the distinct values in the <paramref name="values"/>. You must have pre-allocated <paramref name="distinctValues"/> and <paramref name="counts"/> yourself. The scheme is destructive, because it modifies the arrays within <paramref name="values"/>. </summary> <param name="values">The values we are binning</param> <param name="distinctValues">This working array will be filled with a sorted list of the distinct values detected within <paramref name="values"/></param> <param name="counts">This working array will be filled with a sorted list of the distinct values detected within <paramref name="values"/></param> <returns>The logical length of both <paramref name="distinctValues"/> and <paramref name="counts"/></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BinFinder.FindBinsFromDistinctCounts(System.Double[],System.Int32[],System.Int32,System.Int32,System.Double[]@,System.Int32@)"> <summary> Finds the bins. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BinFinder.IsTrivial(System.Int32[],System.Int32,System.Int32)"> <summary> Check to see if we can "trivialize" this feature, because it would be impossible to split with the indicated minimum examples per leaf. </summary> <param name="distinctCounts">The counts of each distinct bin value</param> <param name="numDistinct">The logical length of <paramref name="distinctCounts"/></param> <param name="minPerLeaf">The minimum examples per leaf we are filtering on</param> <returns>Whether this feature is trivial, that is, it would be impossible to split on it</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.BinFinder.FindBins(Microsoft.ML.Runtime.Data.VBuffer{System.Double}@,System.Int32,System.Int32,System.Double[]@)"> <summary> Finds the bins. </summary> <param name="values">The values we are binning</param> <param name="maxBins">The maximum number of bins to find</param> <param name="minPerLeaf">The minimum number of documents per leaf</param> <param name="binUpperBounds">The calculated upper bound of each bin</param> <returns>Whether finding the bins is successful. If there were NaN values in <paramref name="values"/>, this will return false and the output arrays will be <c>null</c>. Otherwise it will return true.</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IniFileParserInterface.GetFeatureMap"> <summary> Gets the map between feature names and indices </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IniFileParserInterface.GetFeatureEvaluators"> <summary> Gets the list of FeatureEvaluators </summary> <returns></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.Internal.IniFileParserInterface.#ctor(System.IntPtr)"> <summary> Creates an InputExtractor wrapper for a given unmanaged pointer </summary> <param name="pInputExtractor"></param> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IniFileParserInterface.FeatureMap"> <summary> Wraps the functions of InputExtractor that map between raw feature names and indices </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.Internal.IniFileParserInterface.FeatureEvaluator"> <summary> Wraps the Evaluate method of InputExtractor (and associated methods) </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRegressionTrainer.GetDatasetRegressionLabels(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> Gets the regression labels that were stored in the dataset skeleton, or constructs them from the ratings if absent. This returns null if the dataset itself is null. </summary> <param name="set">The dataset</param> <returns>The list of regression targets, or null if <paramref name="set"/> was null</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.FastTreeTweedieTrainer"> <summary> The Tweedie boosting model follows the mathematics established in: Yang, Quan, and Zou. "Insurance Premium Prediction via Gradient Tree-Boosted Tweedie Compound Poisson Models." https://arxiv.org/pdf/1508.06378.pdf </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeTweedieTrainer.GetDatasetRegressionLabels(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> Gets the regression labels that were stored in the dataset skeleton, or constructs them from the ratings if absent. This returns null if the dataset itself is null. </summary> <param name="set">The dataset</param> <returns>The list of regression targets, or null if <paramref name="set"/> was null</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRankingTrainer.CreateStandardTest(Microsoft.ML.Runtime.FastTree.Internal.Dataset)"> <summary> Create standard test for dataset. </summary> <param name="dataset">dataset used for testing</param> <returns>standard test for the dataset</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRankingTrainer.CreateSpecialTrainSetTest"> <summary> Create the special test for train set. </summary> <returns>test for train set</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRankingTrainer.CreateSpecialValidSetTest"> <summary> Create the special test for valid set. </summary> <returns>test for train set</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRankingTrainer.CreateFirstTestSetTest"> <summary> Create the test for the first test set. </summary> <returns>test for the first test set</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeRankingTrainer.GetTestGraphHeader"> <summary> Get the header of test graph </summary> <returns>Test graph header</returns> </member> <member name="F:Microsoft.ML.Runtime.FastTree.TreeArgs.GainConfidenceLevel"> Only consider a gain if its likelihood versus a random choice gain is above a certain value. So 0.95 would mean restricting to gains that have less than a 0.05 change of being generated randomly through choice of a random split. </member> <member name="T:Microsoft.ML.Runtime.FastTree.FastTreeShared"> <summary> FastTreeTrainerBase is generic class and can't have shared object among classes. This class is to provide common for all classes object which we can use for lock purpose. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeTrainerBase`2.GetTestGraphHeader"> <summary> A virtual method that is used to print header of test graph. Appliations that need printing test graph are supposed to override it to print specific test graph header. </summary> <returns> string representation of test graph header </returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeTrainerBase`2.GetTestGraphLine"> <summary> A virtual method that is used to print a single line of test graph. Applications that need printing test graph are supposed to override it to print a specific line of test graph after a new iteration is finished. </summary> <returns> string representation of a line of test graph </returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeTrainerBase`2.ComputeTests"> <summary> A virtual method that is used to compute test results after each iteration is finished. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreeTrainerBase`2.MakeGradientWrapper(Microsoft.ML.Runtime.IChannel)"> <summary> Creates weights wrapping (possibly, trivial) for gradient target values. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.DataConverter.BinUpperBounds"> <summary> The per-feature bin upper bounds. Implementations may differ on when all of the items in this array are initialized to non-null values but it must happen at least no later than immediately after we return from <see cref="M:Microsoft.ML.Runtime.FastTree.DataConverter.GetDataset"/>. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.DataConverter.FeatureMap"> <summary> In the event that any features are filtered, this will contain the feature map, where the indices are the indices of features within the dataset, and the tree as we are learning, and the values are the indices of the features within the original input data. This array is used to "rehydrate" the tree once we finish training, so that the feature indices are once again over the full set of features, as opposed to the subset of features we actually trained on. This can be null in the event that no filtering occurred. </summary> <seealso cref="M:Microsoft.ML.Runtime.FastTree.Internal.Ensemble.RemapFeatures(System.Int32[])"/> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.CalculateBins(Microsoft.ML.Runtime.FastTree.Internal.BinFinder,Microsoft.ML.Runtime.Data.VBuffer{System.Double}@,System.Int32,System.Int32,System.Double[]@,System.Int32[]@,System.Double[]@)"> <summary> Bins and input vector of feature values. </summary> <param name="binFinder">The instead of the bin finder to use</param> <param name="values">The values for one particular feature value across all examples</param> <param name="maxBins">The maximum number of bins to find</param> <param name="minDocsPerLeaf"></param> <param name="distinctValues">The working array of distinct values, a temporary buffer that should be called to multiple invocations of this method, but not meant to be useful to the caller. This method will reallocate the array to a new size if necessary. Passing in null at first is acceptable.</param> <param name="distinctCounts">Similar working array, but for distinct counts</param> <param name="upperBounds">The bin upper bounds, maximum length will be <paramref name="maxBins"/></param> <returns>Whether finding the bins was successful or not. It will be unsuccessful iff <paramref name="values"/> has any missing values. In that event, the out parameters will be left as null.</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.CreateSingletonFlock(Microsoft.ML.Runtime.IChannel,Microsoft.ML.Runtime.Data.VBuffer{System.Double}@,System.Int32[],System.Double[])"> <summary> Create a new feature flock with a given name, values and specified bin bounds. </summary> <param name="ch"></param> <param name="values">The values for this feature, that will be binned.</param> <param name="binnedValues">A working array of length equal to the length of the input feature vector</param> <param name="binUpperBounds">The upper bounds of the binning of this feature.</param> <returns>A derived binned derived feature vector.</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.DiskImpl.ConstructDropSlotRanges(Microsoft.ML.Runtime.Data.ISlotCursor,Microsoft.ML.Runtime.Data.ValueGetter{Microsoft.ML.Runtime.Data.VBuffer{System.Single}},Microsoft.ML.Runtime.Data.VBuffer{System.Single}@)"> <summary> Returns a slot dropper object that has ranges of slots to be dropped, based on an examination of the feature values. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.MemImpl.CreateDatasetSkeleton"> <summary> Create an artificial metadata object to pad the Dataset </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList"> <summary> A mutable list of index,value that may be kept sparse or dense. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList.CountIndicesGT(System.Int32,System.Double)"> <summary> Returns the count of all positions greater than an indicated value. </summary> <param name="length">The limit of indices to check</param> <param name="gtValue">The value against which the greater-than comparison is made</param> <returns>The count of all indices in the range of 0 to <paramref name="length"/> exclusive whose values are greater than <paramref name="gtValue"/></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList.AllIndicesGT(System.Int32,System.Double)"> <summary> Return all indices that are greater than an indicated value. </summary> <param name="lim">The limit of indices to return</param> <param name="gtValue">The value against which the greater-than comparison is made</param> <returns>All indices in the range of 0 to <paramref name="lim"/> exclusive whose values are greater than <paramref name="gtValue"/>, in increasing order</returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList.Binned(System.Double[],System.Int32)"> <summary> An enumerable of the row/bin pair of every non-zero bin row according to the binning passed into this method. </summary> <param name="binUpperBounds">The binning to use for the enumeration</param> <param name="length">The number of rows in this feature</param> <returns>An enumerable that returns a pair of every row-index and binned value, where the row indices are increasing, the binned values are positive</returns> </member> <member name="P:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList.ForwardIndexer.Item(System.Int32,System.Int32)"> <summary> Access the value of a particular feature, at a particular row. </summary> <param name="featureIndex">A feature index, which indexes not the global feature indices, but the index into the subset of features specified at the constructor time</param> <param name="rowIndex">The row index to access, which must be non-decreasing, and must indeed be actually increasing for access on the same feature (e.g., if you have two features, it is OK to access <c>[1, 5]</c>, then <c>[0, 5]</c>, but once this is done you cannot access the same feature at the same position.</param> <returns></returns> </member> <member name="M:Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList.ForwardIndexer.#ctor(Microsoft.ML.Runtime.FastTree.DataConverter.ValuesList[],System.Int32[],System.Int32[]@)"> <summary> Initialize a forward indexer. </summary> <param name="values">Holds the values of the features</param> <param name="features">The array of feature indices this will index</param> <param name="workArray">A possibly shared working array, once used by this forward indexer it should not be used in any previously created forward indexer</param> </member> <member name="P:Microsoft.ML.Runtime.FastTree.ExamplesToFastTreeBins.BinUpperBounds"> <summary> Bin boundaries </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.SaveAsCode(System.IO.TextWriter,Microsoft.ML.Runtime.Data.RoleMappedSchema)"> <summary> write out a C# representation of the ensemble </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.SaveAsText(System.IO.TextWriter,Microsoft.ML.Runtime.Data.RoleMappedSchema)"> <summary> Output the INI model to a given writer </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.SaveAsIni(System.IO.TextWriter,Microsoft.ML.Runtime.Data.RoleMappedSchema,Microsoft.ML.Runtime.Internal.Calibration.ICalibrator)"> <summary> Output the INI model to a given writer </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.AddCalibrationToIni(System.String,Microsoft.ML.Runtime.Internal.Calibration.ICalibrator)"> <summary> Get the calibration summary in INI format </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.GetSummaryInKeyValuePairs(Microsoft.ML.Runtime.Data.RoleMappedSchema)"> <inheritdoc/> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.SaveEnsembleAsCode(System.IO.TextWriter,Microsoft.ML.Runtime.Data.RoleMappedSchema)"> <summary> returns a C# representation of the ensemble </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.SaveTreeAsCode(Microsoft.ML.Runtime.FastTree.Internal.RegressionTree,System.IO.TextWriter,Microsoft.ML.Runtime.Data.VBuffer{Microsoft.ML.Runtime.Data.DvText}@)"> <summary> Convert a single tree to code, called recursively </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.FastTreePredictionWrapper.GetLeaf(System.Int32,Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,System.Collections.Generic.List{System.Int32}@)"> <summary> Returns the leaf node in the requested tree for the given feature vector, and populates 'path' with the list of internal nodes in the path from the root to that leaf. If 'path' is null a new list is initialized. All elements in 'path' are cleared before filling in the current path nodes. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.GamTrainerBase`2"> <summary> Generalized Additive Model Learner. </summary> </member> <member name="F:Microsoft.ML.Runtime.FastTree.GamTrainerBase`2.ArgumentsBase.GainConfidenceLevel"> Only consider a gain if its likelihood versus a random choice gain is above a certain value. So 0.95 would mean restricting to gains that have less than a 0.05 change of being generated randomly through choice of a random split. </member> <member name="M:Microsoft.ML.Runtime.FastTree.GamPredictorBase.GetFeatureContributions(Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,Microsoft.ML.Runtime.Data.VBuffer{System.Single}@,Microsoft.ML.Runtime.Data.BufferBuilder{System.Single}@)"> <summary> Returns a vector of feature contributions for a given example. <paramref name="builder"/> is used as a buffer to accumulate the contributions across trees. If <paramref name="builder"/> is null, it will be created, otherwise it will be reused. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand"> <summary> The GAM model visualization command. Because the data access commands must access private members of <see cref="T:Microsoft.ML.Runtime.FastTree.GamPredictorBase"/>, it is convenient to have the command itself nested within the base predictor class. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.NumFeatures"> <summary> These are the number of input features, as opposed to the number of features used within GAM which may be lower. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.SaveIfNeeded(Microsoft.ML.Runtime.IHost,Microsoft.ML.Runtime.IChannel,System.String)"> <summary> This will write out a file, if needed. In all cases if something is written it will return a version number, with an indication based on sign of whether anything was actually written in this call. </summary> <param name="host">The host from the command</param> <param name="ch">The channel from the command</param> <param name="outFile">The (optionally empty) output file</param> <returns>Returns <c>null</c> if the model file could not be saved because <paramref name="outFile"/> was <c>null</c> or whitespace. Otherwise, if the current version if newer than the last version saved, it will save, and return that version. (In this case, the number is non-negative.) Otherwise, if the current version was the last version saved, then it will return the bitwise not of that version number (in this case, the number is negative).</returns> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.FeatureInfo.UpperBounds"> <summary> The upper bounds of each bin. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.FeatureInfo.BinEffects"> <summary> The amount added to the model for a document falling in a given bin. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.FeatureInfo.DocCounts"> <summary> The number of documents in each bin. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.FeatureInfo.Version"> <summary> The version of the GAM context that has these values. </summary> </member> <member name="P:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Context.FeatureInfo.CategoricalFeatureIndex"> <summary> For features belonging to the same categorical, this value will be the same, Set to -1 for non-categoricals. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.GamPredictorBase.VisualizationCommand.Init(Microsoft.ML.Runtime.IChannel)"> <summary> Attempts to initialize required items, from the input model file. In the event that anything goes wrong, this method will throw. </summary> <param name="ch">The channel</param> <returns>A structure containing essential information about the GAM dataset that enables operations on top of that structure.</returns> </member> <member name="T:Microsoft.ML.Runtime.FastTree.SumupPerformanceCommand"> <summary> This is an internal utility command to measure the performance of the IntArray sumup operation. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.SignatureParallelTrainer"> <summary> Signature of Parallel trainer. </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.FindBestThresholdFromRawArrayFun"> <summary> delegate function. This function is implemented in TLC, and called by TLC++. It will find best threshold from raw histogram data (countByBin, sumTargetsByBin, sumWeightsByBin, numDocsInLeaf, sumTargets, sumWeights) </summary> </member> <member name="T:Microsoft.ML.Runtime.FastTree.IParallelTraining"> <summary> Interface used for parallel training. Mainly contains three parts: 1. interactive with IO: <see href="GetLocalBinConstructionFeatures" />, <see href="SyncGlobalBoundary" />. Data will be partitioned by rows in Data parallel and Voting Parallel. To speed up the find bin process, it let different workers to find bins for different features. Then perform global sync up. In Feature parallel, every machines holds all data, so this is unneeded. 2. interactive with TreeLearner: <see href="InitIteration" />, <see href="CacheHistogram" />, <see href="IsNeedFindLocalBestSplit" />, <see href="IsSkipNonSplittableHistogram" />, <see href="FindGlobalBestSplit" />, <see href="GetGlobalDataCountInLeaf" />, <see href="PerformGlobalSplit" />. A full process is: Use <see href="InitIteration" /> to alter local active features. Use <see href="GetGlobalDataCountInLeaf" /> to check smaller leaf and larger leaf. Use <see href="CacheHistogram" />, <see href="IsNeedFindLocalBestSplit" /> and <see href="IsSkipNonSplittableHistogram" /> to interactive with Feature histograms. Use <see href="FindGlobalBestSplit" /> to sync up global best split Use <see href="PerformGlobalSplit" /> to record global num_data in leaves. 3. interactive with Application : <see href="GlobalMean" />. Output of leaves is calculated by newton step ( - sum(first_order_gradients) / sum(second_order_gradients)). If data is partitioned by row, it needs to a sync up for these sum result. So It needs to call this to get the real output of leaves. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.InitEnvironment"> <summary> Initialize the network connection. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.FinalizeEnvironment"> <summary> Finalize the network. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.InitTreeLearner(Microsoft.ML.Runtime.FastTree.Internal.Dataset,System.Int32,System.Int32,System.Int32@)"> <summary> Initialize once while construct tree learner. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.FinalizeTreeLearner"> <summary> Finalize while tree learner is freed. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.InitIteration(System.Boolean[]@)"> <summary> Initialize every time before training a tree. will alter activeFeatures in Feature parallel. Because it only need to find threshold for part of features in feature parallel. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.FinalizeIteration"> <summary> Finalize after trained one tree. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.CacheHistogram(System.Boolean,System.Int32,System.Int32,Microsoft.ML.Runtime.FastTree.Internal.SufficientStatsBase,System.Boolean)"> <summary> Cache Histogram, it will be used for global aggregate. Only used in Data parallel and Voting Parallel </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.IsNeedFindLocalBestSplit"> <summary> Only return False in Data parallel. Data parallel find best threshold after merged global histograms. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.IsSkipNonSplittableHistogram"> <summary> True if need to skip non-splittable histogram. Only will return False in Voting parallel. That is because local doesn't have global histograms in Voting parallel, So the information about NonSplittable is not correct, and we cannot skip it. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.FindGlobalBestSplit(Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates,Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.LeafSplitCandidates,Microsoft.ML.Runtime.FastTree.FindBestThresholdFromRawArrayFun,Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.SplitInfo[])"> <summary> Find best split among machines. will save result in bestSplits. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.GetGlobalDataCountInLeaf(System.Int32,System.Int32@)"> <summary> Get global num_data on specific leaf. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.PerformGlobalSplit(System.Int32,System.Int32,System.Int32,Microsoft.ML.Runtime.FastTree.Internal.LeastSquaresRegressionTreeLearner.SplitInfo)"> <summary> Used to record the global num_data on leaves. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.GlobalMean(Microsoft.ML.Runtime.FastTree.Internal.Dataset,Microsoft.ML.Runtime.FastTree.Internal.RegressionTree,Microsoft.ML.Runtime.FastTree.Internal.DocumentPartitioning,System.Double[],System.Boolean)"> <summary> Get Global mean on different machines for data partitioning in tree. Used for calculating leaf output value. will return a array this is the mean output of all leaves. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.GetLocalBinConstructionFeatures(System.Int32)"> <summary> Get indices of features that should be find bin in local. After construct local boundary, should call <see href="SyncGlobalBoundary" /> to get boundaries for all features. </summary> </member> <member name="M:Microsoft.ML.Runtime.FastTree.IParallelTraining.SyncGlobalBoundary(System.Int32,System.Int32,System.Double[][])"> <summary> Sync Global feature bucket. used in Data parallel and Voting parallel. Data are partitioned by row. To speed up the Global find bin process, we can let different workers construct Bin Boundary for different features, then perform a global sync up. </summary> </member> <member name="M:Microsoft.ML.Runtime.QuantileStatistics.#ctor(System.Single[],System.Single[],System.Boolean)"> <summary> data array will be modified because of sorting if it is not already sorted yet and this class owns the data. Modifying the data outside will lead to erroneous output by this class </summary> </member> <member name="M:Microsoft.ML.Runtime.QuantileStatistics.GetQuantile(System.Single)"> <summary> There are many ways to estimate quantile. This implementations is based on R-8, SciPy-(1/3,1/3) http://en.wikipedia.org/wiki/Quantile#Estimating_the_quantiles_of_a_population </summary> </member> <member name="T:Microsoft.ML.Runtime.Data.TreeEnsembleFeaturizerBindableMapper"> <summary> A bindable mapper wrapper for tree ensembles, that creates a bound mapper with three outputs: 1. A vector containing the individual tree outputs of the tree ensemble. 2. An indicator vector for the leaves that the feature vector falls on in the tree ensemble. 3. An indicator vector for the internal nodes on the paths that the feature vector falls on in the tree ensemble. </summary> </member> <member name="T:Microsoft.ML.Runtime.Data.TreeEnsembleFeaturizerTransform.ArgumentsForEntryPoint"> <summary> REVIEW: Ideally we should have only one arguments class by using IComponentFactory for the model. For now it probably warrants a REVIEW comment here in case we'd like to merge these two arguments in the future. Also, it might be worthwhile to extract the common arguments to a base class. </summary> </member> </members> </doc> |