org.deeplearning4j.optimize.solvers

Class StochasticGradientDescent

java.lang.Object

org.deeplearning4j.optimize.solvers.BaseOptimizer

org.deeplearning4j.optimize.solvers.StochasticGradientDescent

All Implemented Interfaces:

Serializable, ConvexOptimizer

public class StochasticGradientDescent
extends BaseOptimizer

Stochastic Gradient Descent Standard fix step size No line search

Author:

Adam Gibson

See Also:

Serialized Form

Field Summary

Fields inherited from class org.deeplearning4j.optimize.solvers.BaseOptimizer

accumulator, computationGraphUpdater, conf, GRADIENT_KEY, lineMaximizer, log, model, oldScore, PARAMS_KEY, score, SCORE_KEY, SEARCH_DIR, searchState, step, stepFunction, stepMax, terminationConditions, trainingListeners, updater

Constructor Summary

Constructors

Constructor and Description
StochasticGradientDescent(NeuralNetConfiguration conf, StepFunction stepFunction, Collection<TrainingListener> trainingListeners, Collection<TerminationCondition> terminationConditions, Model model)
StochasticGradientDescent(NeuralNetConfiguration conf, StepFunction stepFunction, Collection<TrainingListener> trainingListeners, Model model)

Method Summary

Modifier and Type	Method and Description
boolean	optimize(LayerWorkspaceMgr workspaceMgr) Optimize call.
void	postStep(org.nd4j.linalg.api.ndarray.INDArray gradient) Post step to update searchDirection with new gradient and parameter information
void	preProcessLine() Pre preProcess to setup initial searchDirection approximation

Methods inherited from class org.deeplearning4j.optimize.solvers.BaseOptimizer

applyConstraints, batchSize, checkTerminalConditions, getComputationGraphUpdater, getConf, getDefaultStepFunctionForOptimizer, getEpochCount, getGradientsAccumulator, getIterationCount, getUpdater, gradientAndScore, incrementIterationCount, postFirstStep, score, setBatchSize, setGradientsAccumulator, setListeners, setUpdater, setUpdaterComputationGraph, setupSearchState, updateGradientAccordingToParams

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.deeplearning4j.optimize.api.ConvexOptimizer

getStepFunction

Constructor Detail

StochasticGradientDescent

public StochasticGradientDescent(NeuralNetConfiguration conf,
                                 StepFunction stepFunction,
                                 Collection<TrainingListener> trainingListeners,
                                 Model model)

StochasticGradientDescent

public StochasticGradientDescent(NeuralNetConfiguration conf,
                                 StepFunction stepFunction,
                                 Collection<TrainingListener> trainingListeners,
                                 Collection<TerminationCondition> terminationConditions,
                                 Model model)

Method Detail

optimize

public boolean optimize(LayerWorkspaceMgr workspaceMgr)

Description copied from class: BaseOptimizer

Optimize call. This runs the optimizer.

Specified by:

optimize in interface ConvexOptimizer

Overrides:

optimize in class BaseOptimizer

Returns:

whether it converged or not

preProcessLine

public void preProcessLine()

Description copied from class: BaseOptimizer

Pre preProcess to setup initial searchDirection approximation

Specified by:

preProcessLine in interface ConvexOptimizer

Overrides:

preProcessLine in class BaseOptimizer

postStep

public void postStep(org.nd4j.linalg.api.ndarray.INDArray gradient)

Description copied from class: BaseOptimizer

Post step to update searchDirection with new gradient and parameter information

Specified by:

postStep in interface ConvexOptimizer

Overrides:

postStep in class BaseOptimizer