Package org.apache.sysds.runtime.controlprogram.paramserv.dp

Class SubsampleToMinFederatedScheme

java.lang.Object

org.apache.sysds.runtime.controlprogram.paramserv.dp.DataPartitionFederatedScheme

org.apache.sysds.runtime.controlprogram.paramserv.dp.SubsampleToMinFederatedScheme

public class SubsampleToMinFederatedScheme
extends DataPartitionFederatedScheme

Subsample to Min Federated scheme When the parameter server runs in federated mode it cannot pull in the data which is already on the workers. Therefore, a UDF is sent to manipulate the data locally. In this case the global minimum number of examples is taken and the worker subsamples data to match that number of examples. The subsampling is done by multiplying with a Permutation Matrix with a global seed. Then all entries in the federation map of the input matrix are separated into MatrixObjects and returned as a list. Only supports row federated matrices atm.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.sysds.runtime.controlprogram.paramserv.dp.DataPartitionFederatedScheme

DataPartitionFederatedScheme.BalanceMetrics, DataPartitionFederatedScheme.Result

Constructor Summary

Constructors

Constructor	Description
SubsampleToMinFederatedScheme()

Method Summary

Modifier and Type	Method	Description
DataPartitionFederatedScheme.Result	partition(MatrixObject features, MatrixObject labels, int seed)

Methods inherited from class java.lang.Object

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

Constructor Detail

SubsampleToMinFederatedScheme

public SubsampleToMinFederatedScheme()

Method Detail

partition

public DataPartitionFederatedScheme.Result partition(MatrixObject features,
                                                     MatrixObject labels,
                                                     int seed)

Specified by:

partition in class DataPartitionFederatedScheme