Package org.tribuo.math.optimisers

Provides implementations of StochasticGradientOptimiser.

See: Description

Class Summary

Class	Description
AdaDelta	An implementation of the AdaDelta gradient optimiser.
AdaGrad	An implementation of the AdaGrad gradient optimiser.
AdaGradRDA	An implementation of the AdaGrad gradient optimiser with regularized dual averaging.
Adam	An implementation of the Adam gradient optimiser.
GradientOptimiserOptions	CLI options for configuring a gradient optimiser.
ParameterAveraging	Averages the parameters across a gradient run.
Pegasos	An implementation of the Pegasos gradient optimiser used primarily for solving the SVM problem.
RMSProp	An implementation of the RMSProp gradient optimiser.
SGD	An implementation of single learning rate SGD and optionally momentum.

Enum Summary

Enum	Description
GradientOptimiserOptions.StochasticGradientOptimiserType	Type of the gradient optimisers available in CLIs.
SGD.Momentum	Momentum types.

Package org.tribuo.math.optimisers Description

Provides implementations of StochasticGradientOptimiser.

Has implementations of SGD using a variety of simple learning rate tempering systems, along with AdaGrad, Adam, AdaDelta, RMSProp and Pegasos.

Also provides ParameterAveraging which wraps another StochasticGradientOptimiser and averages the learned parameters across the gradient descent run. This is usually used for convex problems, for non-convex ones your milage may vary.