org.apache.ignite.ml.dataset.impl.bootstrapping

Class BootstrappedDatasetBuilder<K,V>

java.lang.Object

org.apache.ignite.ml.dataset.impl.bootstrapping.BootstrappedDatasetBuilder<K,V>

Type Parameters:

K - Type of a key in upstream data.

V - Type of a value in upstream data.

All Implemented Interfaces:

Serializable, PartitionDataBuilder<K,V,EmptyContext,BootstrappedDatasetPartition>

public class BootstrappedDatasetBuilder<K,V>
extends Object
implements PartitionDataBuilder<K,V,EmptyContext,BootstrappedDatasetPartition>

Builder for bootstrapped dataset. Bootstrapped dataset consist of several subsamples created in according to random sampling with replacements selection of vectors from original dataset. This realization uses BootstrappedVector containing each vector from original sample with counters of repetitions for each subsample. As heuristic this implementation uses Poisson Distribution for generating counter values.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
BootstrappedDatasetBuilder(Preprocessor<K,V> preprocessor, int samplesCnt, double subsampleSize) Creates an instance of BootstrappedDatasetBuilder.

Method Summary

Modifier and Type	Method and Description
BootstrappedDatasetPartition	build(LearningEnvironment env, Iterator<UpstreamEntry<K,V>> upstreamData, long upstreamDataSize, EmptyContext ctx) Builds a new partition data from a partition upstream data and partition context.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.apache.ignite.ml.dataset.PartitionDataBuilder

andThen, build

Constructor Detail

BootstrappedDatasetBuilder

public BootstrappedDatasetBuilder(Preprocessor<K,V> preprocessor,
                                  int samplesCnt,
                                  double subsampleSize)

Creates an instance of BootstrappedDatasetBuilder.

Parameters:

preprocessor - Mapper of upstream entries into LabeledVector.

samplesCnt - Samples count.

subsampleSize - Subsample size.

Method Detail

build

public BootstrappedDatasetPartition build(LearningEnvironment env,
                                          Iterator<UpstreamEntry<K,V>> upstreamData,
                                          long upstreamDataSize,
                                          EmptyContext ctx)

Builds a new partition data from a partition upstream data and partition context. Important: there is no guarantee that there will be no more than one UpstreamEntry with given key, UpstreamEntry should be thought rather as a container saving all data from upstream, but omitting uniqueness constraint. This constraint is omitted to allow upstream data transformers in DatasetBuilder replicating entries. For example it can be useful for bootstrapping.

Specified by:

build in interface PartitionDataBuilder<K,V,EmptyContext,BootstrappedDatasetPartition>

Parameters:

env - Learning environment.

upstreamData - Partition upstream data.

upstreamDataSize - Partition upstream data size.

ctx - Partition context.

Returns:

Partition data.