MachineLearning::Layer Class Reference

A class representing a single layer within a neural network. More...

#include <Layer.h>

Public Member Functions
	AZ_TYPE_INFO (Layer, "{FB91E0A7-86C0-4431-83A8-04F8D8E1C9E2}")
	Layer (Layer &&)=default
	Layer (const Layer &)=default
	Layer (ActivationFunctions activationFunction, AZStd::size_t activationDimensionality, AZStd::size_t layerDimensionality)
Layer &	operator= (Layer &&)=default
Layer &	operator= (const Layer &)=default
const AZ::VectorN &	Forward (LayerInferenceData &inferenceData, const AZ::VectorN &activations)
	Performs a basic forward pass on this layer, outputs are stored in m_output.
void	AccumulateGradients (AZStd::size_t samples, LayerTrainingData &trainingData, LayerInferenceData &inferenceData, const AZ::VectorN &expected)
void	ApplyGradients (LayerTrainingData &trainingData, float learningRate)
	Applies the current gradient values to the layers weights and biases and resets the gradient values for a new accumulation pass.
bool	Serialize (AzNetworking::ISerializer &serializer)
AZStd::size_t	EstimateSerializeSize () const
	Returns the estimated size required to serialize this layer.
void	OnSizesChanged ()
	Updates layer internals for it's requested dimensionalities.

Static Public Member Functions
static void	Reflect (AZ::ReflectContext *context)

Public Attributes
AZStd::size_t	m_inputSize = 0
AZStd::size_t	m_outputSize = 0
AZ::MatrixMxN	m_weights
AZ::VectorN	m_biases
ActivationFunctions	m_activationFunction = ActivationFunctions::ReLU

Detailed Description

A class representing a single layer within a neural network.

Member Function Documentation

AccumulateGradients()

void MachineLearning::Layer::AccumulateGradients	(	AZStd::size_t	samples,
		LayerTrainingData &	trainingData,
		LayerInferenceData &	inferenceData,
		const AZ::VectorN &	expected
	)

Performs a gradient computation against the provided expected output using the provided gradients from the previous layer. This method presumes that we've completed a forward pass immediately prior to fill all the relevant vectors

Reflect()

static void MachineLearning::Layer::Reflect ( AZ::ReflectContext *  context )

static

AzCore Reflection.

Parameters

context

reflection context

Serialize()

bool MachineLearning::Layer::Serialize

(

AzNetworking::ISerializer &

serializer

)

Base serialize method for all serializable structures or classes to implement.

Parameters

serializer

ISerializer instance to use for serialization

Returns

boolean true for success, false for serialization failure

---

The documentation for this class was generated from the following file:

• Gems/MachineLearning/Code/Source/Models/Layer.h