DataPredict

API Reference - Optimizers - BaseOptimizer

BaseOptimizer is a base for all optimizers.

Constructors

new()

Creates a new base optimizer object.

BaseOptimizer.new(optimizerName: string): BaseOptimizerObject

Parameters

• optimizerName: The optimizer name that is stored in base optimizer.

Returns:

• OptimizerObject: The generated optimizer object.

Functions

calculate()

Returns a modified cost function derivatives.

BaseOptimizer:calculate(learningRate: number, costFunctionDerivatives: matrix): matrix

Parameters:

• learningRate: The learning rate used by a model.

• costFunctionDerivatives: The cost function derivatives calculated by a model.

Returns:

• costFunctionDerivatives: The modified cost function derivatives that is to be used by a model.

getOptimizerName()

Gets the optimizer’s name from the base optimizer.

BaseOptimizer:getOptimizerName()

Returns:

• optimizerName: The optimizer name that is stored in base optimizer.

setCalculateFunction()

Sets a calculate function for the base optimizer.

BaseOptimizer:setCalculateFunction(calculateFunction: Function)

Parameters:

• The calculate function to be used by the base optimizer when calculate() function is called.

setLearningRateValueScheduler()

Sets a value scheduler for the learning rate.

BaseOptimizer:setLearningRateScheduler(LearningRateValueScheduler: ValueSchedulerObject)

Parameters:

LearningRateValueScheduler: The value scheduler object to be used by the learning rate.

getLearningRateValueScheduler()

Gets the value scheduler for the learning rate.

BaseOptimizer:getLearningRateScheduler(): ValueSchedulerObject

Returns:

LearningRateValueScheduler: The value scheduler object that was used by the learning rate.

getOptimizerInternalParameters()

Gets the optimizer’s internal parameters from the base optimizer.

BaseOptimizer:getOptimizerInternalParameters(doNotDeepCopy: boolean): {}

Parameters:

• doNotDeepCopy: Set whether or not to deep copy the optimizer internal parameters.

Returns:

• optimizerInternalParameters: A matrix/table containing optimizer internal parameters fetched from the base optimizer.

setOptimizerInternalParameters()

Sets the optimizer’s internal parameters from the base optimizer.

BaseOptimizer:setOptimizerInternalParameters(optimizerInternalParameters: {}, doNotDeepCopy: boolean)

Parameters:

• optimizerInternalParameters: A matrix/table containing optimizer internal parameters that will be used by the base optimizer.

• doNotDeepCopy: Set whether or not to deep copy the optimizer internal parameters.

reset()

Reset optimizer’s stored values (excluding the parameters).

BaseOptimizer:reset()