org.apache.spark.mllib.clustering

Class OnlineLDAOptimizer

Object

org.apache.spark.mllib.clustering.OnlineLDAOptimizer

All Implemented Interfaces:

Logging, LDAOptimizer

public final class OnlineLDAOptimizer
extends Object
implements LDAOptimizer, Logging

:: DeveloperApi ::

An online optimizer for LDA. The Optimizer implements the Online variational Bayes LDA algorithm, which processes a subset of the corpus on each iteration, and updates the term-topic distribution adaptively.

Original Online LDA paper: Hoffman, Blei and Bach, "Online Learning for Latent Dirichlet Allocation." NIPS, 2010.

Constructor Summary

Constructors

Constructor and Description
OnlineLDAOptimizer()

Method Summary

Modifier and Type	Method and Description
double	getKappa() Learning rate: exponential decay rate
double	getMiniBatchFraction() Mini-batch fraction, which sets the fraction of document sampled and used in each iteration
boolean	getOptimizeDocConcentration() Optimize docConcentration, indicates whether docConcentration (Dirichlet parameter for document-topic distribution) will be optimized during training.
double	getTau0() A (positive) learning parameter that downweights early iterations.
OnlineLDAOptimizer	setKappa(double kappa) Learning rate: exponential decay rate---should be between (0.5, 1.0] to guarantee asymptotic convergence.
OnlineLDAOptimizer	setMiniBatchFraction(double miniBatchFraction) Mini-batch fraction in (0, 1], which sets the fraction of document sampled and used in each iteration.
OnlineLDAOptimizer	setOptimizeDocConcentration(boolean optimizeDocConcentration) Sets whether to optimize docConcentration parameter during training.
OnlineLDAOptimizer	setTau0(double tau0) A (positive) learning parameter that downweights early iterations.

Methods inherited from class Object

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

Methods inherited from interface org.apache.spark.internal.Logging

initializeLogging, initializeLogIfNecessary, initializeLogIfNecessary, isTraceEnabled, log_, log, logDebug, logDebug, logError, logError, logInfo, logInfo, logName, logTrace, logTrace, logWarning, logWarning

Constructor Detail

OnlineLDAOptimizer

public OnlineLDAOptimizer()

Method Detail

getTau0

public double getTau0()

A (positive) learning parameter that downweights early iterations. Larger values make early iterations count less.

Returns:

(undocumented)

setTau0

public OnlineLDAOptimizer setTau0(double tau0)

A (positive) learning parameter that downweights early iterations. Larger values make early iterations count less. Default: 1024, following the original Online LDA paper.

Parameters:

tau0 - (undocumented)

Returns:

(undocumented)

getKappa

public double getKappa()

Learning rate: exponential decay rate

Returns:

(undocumented)

setKappa

public OnlineLDAOptimizer setKappa(double kappa)

Learning rate: exponential decay rate---should be between (0.5, 1.0] to guarantee asymptotic convergence. Default: 0.51, based on the original Online LDA paper.

Parameters:

kappa - (undocumented)

Returns:

(undocumented)

getMiniBatchFraction

public double getMiniBatchFraction()

Mini-batch fraction, which sets the fraction of document sampled and used in each iteration

Returns:

(undocumented)

setMiniBatchFraction

public OnlineLDAOptimizer setMiniBatchFraction(double miniBatchFraction)

Mini-batch fraction in (0, 1], which sets the fraction of document sampled and used in each iteration.

Parameters:

miniBatchFraction - (undocumented)

Returns:

(undocumented)

Note:

This should be adjusted in synch with LDA.setMaxIterations() so the entire corpus is used. Specifically, set both so that maxIterations * miniBatchFraction is at least 1.

Default: 0.05, i.e., 5% of total documents.

getOptimizeDocConcentration

public boolean getOptimizeDocConcentration()

Optimize docConcentration, indicates whether docConcentration (Dirichlet parameter for document-topic distribution) will be optimized during training.

Returns:

(undocumented)

setOptimizeDocConcentration

public OnlineLDAOptimizer setOptimizeDocConcentration(boolean optimizeDocConcentration)

Sets whether to optimize docConcentration parameter during training.

Default: false

Parameters:

optimizeDocConcentration - (undocumented)

Returns:

(undocumented)