Package org.jetbrains.kotlinx.dl.api.core.optimizer

Types

AdaDelta

class AdaDelta(learningRate: Float, rho: Float, epsilon: Float, clipGradient: ClipGradientAction) : Optimizer

Adadelta optimizer.

AdaGrad

class AdaGrad(learningRate: Float, initialAccumulatorValue: Float, clipGradient: ClipGradientAction) : Optimizer

Adagrad optimizer.

AdaGradDA

class AdaGradDA(learningRate: Float, initialAccumulatorValue: Float, l1Strength: Float, l2Strength: Float, clipGradient: ClipGradientAction) : Optimizer

Adagrad Dual Averaging algorithm for sparse linear models.

Adam

class Adam(learningRate: Float, beta1: Float, beta2: Float, epsilon: Float, useNesterov: Boolean, clipGradient: ClipGradientAction) : Optimizer

Adam optimizer.

Adamax

class Adamax(learningRate: Float, beta1: Float, beta2: Float, epsilon: Float, clipGradient: ClipGradientAction) : Optimizer

Adamax optimizer from Adam paper's Section 7.

ClipGradientAction

Base abstract class for approaches to clip gradient values from step to step in optimizer.

ClipGradientByNorm

Clips gradient with pre-defined clipNormValue norm.

ClipGradientByValue

Clips gradient value to clipValue if gradient value more than clipValue and to -clipValue if gradient value less than -clipValue.

Ftrl

class Ftrl(learningRate: Float, l1RegularizationStrength: Float, l2RegularizationStrength: Float, learningRatePower: Float, l2ShrinkageRegularizationStrength: Float, initialAccumulatorValue: Float, clipGradient: ClipGradientAction) : Optimizer

Optimizer that implements the FTRL algorithm.

Momentum

class Momentum(learningRate: Float, momentum: Float, useNesterov: Boolean, clipGradient: ClipGradientAction) : Optimizer

Improved version of SGD optimizer.

NoClipGradient

No gradient clipping. Gradients go forward without any changes.

Optimizer

abstract class Optimizer(clipGradient: ClipGradientAction)

Base class for all optimizers.

RMSProp

class RMSProp(learningRate: Float, decay: Float, momentum: Float, epsilon: Float, centered: Boolean, clipGradient: ClipGradientAction) : Optimizer

RMSProp optimizer.

SGD

Stochastic gradient descent optimizer.