LightGBM::OrderedBin Class Reference

Interface for ordered bin data. efficient for construct histogram, especially for sparse bin There are 2 advantages by using ordered bin. More...

#include <bin.h>

Inheritance diagram for LightGBM::OrderedBin:

Public Member Functions
virtual	~OrderedBin ()
	virtual destructor
virtual void	Init (const char *used_indices, data_size_t num_leaves)=0
	Initialization logic.
virtual void	ConstructHistogram (int leaf, const score_t *gradients, const score_t *hessians, HistogramBinEntry *out) const =0
	Construct histogram by using this bin Note: Unlike Bin, OrderedBin doesn't use ordered gradients and ordered hessians. Because it is hard to know the relative index in one leaf for sparse bin, since we skipped zero bins.
virtual void	ConstructHistogram (int leaf, const score_t *gradients, HistogramBinEntry *out) const =0
	Construct histogram by using this bin Note: Unlike Bin, OrderedBin doesn't use ordered gradients and ordered hessians. Because it is hard to know the relative index in one leaf for sparse bin, since we skipped zero bins.
virtual void	Split (int leaf, int right_leaf, const char *is_in_leaf, char mark)=0
	Split current bin, and perform re-order by leaf.
virtual data_size_t	NonZeroCount (int leaf) const =0

Detailed Description

Interface for ordered bin data. efficient for construct histogram, especially for sparse bin There are 2 advantages by using ordered bin.

1. group the data by leafs to improve the cache hit.
2. only store the non-zero bin, which can speed up the histogram consturction for sparse features. However it brings additional cost: it need re-order the bins after every split, which will cost much for dense feature. So we only using ordered bin for sparse situations.

Member Function Documentation

ConstructHistogram() [1/2]

virtual void LightGBM::OrderedBin::ConstructHistogram ( int  leaf, const score_t *  gradients, const score_t *  hessians, HistogramBinEntry *  out  ) const

pure virtual

Construct histogram by using this bin Note: Unlike Bin, OrderedBin doesn't use ordered gradients and ordered hessians. Because it is hard to know the relative index in one leaf for sparse bin, since we skipped zero bins.

Parameters

leaf	Using which leaf's data to construct
gradients	Gradients, Note:non-oredered by leaf
hessians	Hessians, Note:non-oredered by leaf
out	Output Result

Implemented in LightGBM::OrderedSparseBin< VAL_T >.

ConstructHistogram() [2/2]

virtual void LightGBM::OrderedBin::ConstructHistogram ( int  leaf, const score_t *  gradients, HistogramBinEntry *  out  ) const

pure virtual

Construct histogram by using this bin Note: Unlike Bin, OrderedBin doesn't use ordered gradients and ordered hessians. Because it is hard to know the relative index in one leaf for sparse bin, since we skipped zero bins.

Parameters

leaf	Using which leaf's data to construct
gradients	Gradients, Note:non-oredered by leaf
out	Output Result

Implemented in LightGBM::OrderedSparseBin< VAL_T >.

Init()

virtual void LightGBM::OrderedBin::Init ( const char *  used_indices, data_size_t  num_leaves  )

pure virtual

Initialization logic.

Parameters

used_indices	If used_indices.size() == 0 means using all data, otherwise, used_indices[i] == true means i-th data is used (this logic was build for bagging logic)
num_leaves	Number of leaves on this iteration

Split()

virtual void LightGBM::OrderedBin::Split ( int  leaf, int  right_leaf, const char *  is_in_leaf, char  mark  )

pure virtual

Split current bin, and perform re-order by leaf.

Parameters

leaf	Using which leaf's to split
right_leaf	The new leaf index after perform this split
is_in_leaf	is_in_leaf[i] == mark means the i-th data will be on left leaf after split
mark	is_in_leaf[i] == mark means the i-th data will be on left leaf after split

Implemented in LightGBM::OrderedSparseBin< VAL_T >.

---

The documentation for this class was generated from the following file:

• External/LightGBM_2.2.3/LightGBM-2.2.3/include/LightGBM/bin.h