FlexNN::Layer Class Reference

Represents a single layer in a neural network. More...

#include <Layer.h>

Public Member Functions
	Layer (int inputSize, int outputSize, const std::string &activationFunction="relu")
	Constructor for the Layer class.
Eigen::MatrixXd	getWeights () const
	Getters for weights.
Eigen::VectorXd	getBiases () const
	Getters for biases.
void	updateWeights (const Eigen::MatrixXd &dW, const Eigen::VectorXd &db, double learningRate)
	Update weights and biases.
std::pair< Eigen::MatrixXd, Eigen::MatrixXd >	forward (const Eigen::MatrixXd &input)
	Forward pass through the layer.
Eigen::MatrixXd	backward (const Eigen::MatrixXd &nextW, const Eigen::MatrixXd &nextdZ, const Eigen::MatrixXd &currZ)
	Backward pass through the layer.

Detailed Description

Represents a single layer in a neural network.

The Layer class encapsulates the properties and methods required for a neural network layer, including weights, biases, forward and backward passes, and weight updates.

This class is designed to be flexible and can be used with different activation functions. It supports both relu and softmax activation functions by default, but can be extended to include others.

Constructor & Destructor Documentation

Layer()

FlexNN::Layer::Layer ( int  inputSize, int  outputSize, const std::string &  activationFunction = "relu"  )

inline

Constructor for the Layer class.

Initializes the layer with random weights and biases.

Parameters

inputSize	The size of the input to this layer.
outputSize	The size of the output from this layer (also the number of neurons of this layer).
activationFunction	The activation function to be used in this layer (default is "relu").

Note

If this is the last layer, the activation function should be "softmax".

Member Function Documentation

backward()

Eigen::MatrixXd FlexNN::Layer::backward	(	const Eigen::MatrixXd &	nextW,
		const Eigen::MatrixXd &	nextdZ,
		const Eigen::MatrixXd &	currZ
	)

Backward pass through the layer.

This method computes the gradient of the loss with respect to the inputs of this layer given the gradients from the next layer.

Parameters

nextW	The weights of the next layer.
nextdZ	The gradients from the next layer.
currZ	The linear output (Z) of this layer.

Returns

The gradient of the loss with respect to the inputs of this layer (dZ).

forward()

std::pair< Eigen::MatrixXd, Eigen::MatrixXd > FlexNN::Layer::forward

(

const Eigen::MatrixXd &

input

)

Forward pass through the layer.

This method computes the output of the layer given an input matrix. It applies the activation function to the linear combination of inputs and weights.

Parameters

input

The input data for the forward pass.

Returns

A pair containing the linear output (Z) and the activated output (A).

getBiases()

Eigen::VectorXd FlexNN::Layer::getBiases ( ) const

inline

Getters for biases.

These methods return the biases of the layer.

Returns

Eigen::VectorXd The biases of the layer.

getWeights()

Eigen::MatrixXd FlexNN::Layer::getWeights ( ) const

inline

Getters for weights.

These methods return the weights of the layer.

Returns

Eigen::MatrixXd The weights of the layer.

updateWeights()

void FlexNN::Layer::updateWeights ( const Eigen::MatrixXd &  dW, const Eigen::VectorXd &  db, double  learningRate  )

inline

Update weights and biases.

This method updates the weights and biases of the layer using the provided gradients and a specified learning rate.

Parameters

dW	The gradient of the weights.
db	The gradient of the biases.
learningRate	The learning rate for updating the weights and biases.

---

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

• include/Layer.h
• lib/Layer.cpp