chisei::NeuralNetwork Class Reference

Represents a fully connected feedforward neural network. More...

#include <neural_network.hpp>

Public Member Functions
	NeuralNetwork (const std::vector< size_t > &_layers, std::function< double(double)> _activation, std::function< double(double)> _activation_derivative)
	Constructs a neural network with the specified layers and activation functions.
	NeuralNetwork (const NeuralNetwork &other)
	Copy constructor.
	~NeuralNetwork ()
	Destructor for the neural network.
NeuralNetwork &	operator= (NeuralNetwork &&other) noexcept
	Move assignment operator.
std::vector< double >	predict (const std::vector< double > &input)
	Predicts the output for a given input vector.
void	train (const std::vector< std::vector< double > > &inputs, const std::vector< std::vector< double > > &targets, double learning_rate=0.1, int epochs=10000)
	Trains the neural network using the provided training data.
double	compute_mse_loss (const std::vector< double > &prediction, const std::vector< double > &target)
	Computes the mean squared error (MSE) loss.
std::vector< double >	compute_output_gradient (const std::vector< double > &prediction, const std::vector< double > &target)
	Computes the gradient of the loss with respect to the output layer.
double	compute_accuracy (const std::vector< std::vector< double > > &inputs, const std::vector< std::vector< double > > &targets)
	Computes the accuracy of the network on a dataset.
bool	is_correct_prediction (const std::vector< double > &prediction, const std::vector< double > &target)
	Determines whether a prediction is correct based on a target.
void	save_model (const std::string &filename)
	Saves the current state of the neural network to a file.

Static Public Member Functions
static NeuralNetwork	loadFromModel (const std::string &filename)
	Loads a neural network from a saved model file.

Detailed Description

This class provides methods for creating, training, and using neural networks. It supports:

• Customizable activation functions.
• Training via backpropagation with mean squared error (MSE) loss.
• Saving and loading models to/from files.

Definition at line 64 of file neural_network.hpp.

Constructor & Destructor Documentation

NeuralNetwork() [1/2]

chisei::NeuralNetwork::NeuralNetwork	(	const std::vector< size_t > &	_layers,
		std::function< double(double)>	_activation,
		std::function< double(double)>	_activation_derivative
	)

Parameters

_layers	A vector specifying the number of neurons in each layer.
_activation	The activation function to use in the network.
_activation_derivative	The derivative of the activation function.

NeuralNetwork() [2/2]

chisei::NeuralNetwork::NeuralNetwork

(

const NeuralNetwork &

other

)

Creates a deep copy of the given neural network.

Parameters

other

The neural network to copy.

~NeuralNetwork()

chisei::NeuralNetwork::~NeuralNetwork

(

)

Member Function Documentation

compute_accuracy()

double chisei::NeuralNetwork::compute_accuracy	(	const std::vector< std::vector< double > > &	inputs,
		const std::vector< std::vector< double > > &	targets
	)

Parameters

inputs	The input data.
targets	The expected outputs.

Returns

The accuracy as a percentage (0.0 to 100.0).

compute_mse_loss()

double chisei::NeuralNetwork::compute_mse_loss	(	const std::vector< double > &	prediction,
		const std::vector< double > &	target
	)

Parameters

prediction	The predicted output vector.
target	The expected output vector.

Returns

The computed MSE loss.

compute_output_gradient()

std::vector< double > chisei::NeuralNetwork::compute_output_gradient	(	const std::vector< double > &	prediction,
		const std::vector< double > &	target
	)

Parameters

prediction	The predicted output vector.
target	The expected output vector.

Returns

The gradient vector.

is_correct_prediction()

bool chisei::NeuralNetwork::is_correct_prediction	(	const std::vector< double > &	prediction,
		const std::vector< double > &	target
	)

Parameters

prediction	The predicted output vector.
target	The expected output vector.

Returns

True if the prediction is correct; otherwise, false.

loadFromModel()

static NeuralNetwork chisei::NeuralNetwork::loadFromModel ( const std::string &  filename )

static

Parameters

filename

The name of the file to load the model from.

Returns

A NeuralNetwork object initialized from the file data.

Exceptions

std::ios_base::failure

if the file cannot be read or is malformed.

operator=()

NeuralNetwork & chisei::NeuralNetwork::operator= ( NeuralNetwork &&  other )

noexcept

Moves the contents of another neural network into this instance.

Parameters

other

The neural network to move from.

Returns

A reference to the current instance.

predict()

std::vector< double > chisei::NeuralNetwork::predict

(

const std::vector< double > &

input

)

Performs a forward pass through the network to compute the output.

Parameters

input

The input vector.

Returns

The output vector.

save_model()

void chisei::NeuralNetwork::save_model

(

const std::string &

filename

)

Parameters

filename

The name of the file to save the model to.

Exceptions

std::ios_base::failure

if the file cannot be written.

train()

void chisei::NeuralNetwork::train	(	const std::vector< std::vector< double > > &	inputs,
		const std::vector< std::vector< double > > &	targets,
		double	learning_rate = 0.1,
		int	epochs = 10000
	)

Uses backpropagation and gradient descent to minimize the loss function.

Parameters

inputs	The training input data.
targets	The expected output data corresponding to the inputs.
learning_rate	The learning rate for gradient descent (default = 0.1).
epochs	The number of training iterations (default = 10,000).

---

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

• include/chisei/neural_network.hpp