Adam

data class Adam(val alpha: Double = 0.001, val beta1: Double = 0.9, val beta2: Double = 0.999, val epsilon: Double = 1.0E-8, val statistics: Statistics) : MultiPassOptimizer, Statistics

Adam optimizer. Based on research paper: https://arxiv.org/pdf/1412.6980

Constructors

constructor(alpha: Double = 0.001, beta1: Double = 0.9, beta2: Double = 0.999, epsilon: Double = 1.0E-8, statistics: Statistics)

val alpha: Double = 0.001

val beta1: Double = 0.9

val beta2: Double = 0.999

open override val epochCount: Int

val epsilon: Double = 1.0E-8

Lists the measurements taken by this object.

open override fun batch(cases: List<Exercise>): List<List<Exercise>>

Creates training batches out of the given cases.

Collects all measures into a single immutable object

open override fun data(key: Measurement<Float>): DataList<Float, Float>

open override fun passAgain(): Boolean

Returns true if the optimizer should perform another pass over the training data.

open override fun update(step: Int, epoch: Int, layer: Layer, weightGradients: Tensor, biasGradients: Tensor)

Updates the parameters of the model based on the outputs computed during the forward pass.