Column sampling at individual nodes (splits). (#3971)

* Column sampling at individual nodes (splits).

* Documented colsample_bynode parameter.

- also updated documentation for colsample_by* parameters

* Updated documentation.

* GetFeatureSet() returns shared pointer to std::vector.

* Sync sampled columns across multiple processes.
This commit is contained in:
Andy Adinets
2018-12-14 15:37:35 +01:00
committed by Jiaming Yuan
parent e0a279114e
commit 42bf90eb8f
8 changed files with 140 additions and 80 deletions

View File

@@ -227,6 +227,7 @@ TEST(GpuHist, EvaluateSplits) {
TrainParam param;
param.max_depth = 1;
param.n_gpus = 1;
param.colsample_bynode = 1;
param.colsample_bylevel = 1;
param.colsample_bytree = 1;
param.min_child_weight = 0.01;
@@ -284,6 +285,7 @@ TEST(GpuHist, EvaluateSplits) {
hist_maker.param_ = param;
hist_maker.shards_.push_back(std::move(shard));
hist_maker.column_sampler_.Init(n_cols,
param.colsample_bynode,
param.colsample_bylevel,
param.colsample_bytree,
false);