Jyn Class Documentation Overview The Jyn class represents a neural network that can be trained using backpropagation. It manages the network's architecture, datasets, forward propagation, and gradient descent optimization.
Fields ArrayList architecture: The structure of the neural network, comprising its layers. ArrayList layers: A list of integers representing the size of each layer. ArrayList<ArrayList> dataset: The dataset used for training the neural network. JMatrix output: The output produced by the network after forward propagation. ArrayList<ArrayList> nn: The list of weights and biases for the network. ArrayList<ArrayList> gradients: The calculated gradients for the weights and biases. ArrayList<Function<JMatrix, JMatrix>> activation_functions: The activation functions for each layer. ArrayList<Function<JMatrix, JMatrix>> activation_derv_functions: The derivatives of the activation functions for each layer. Constructor Jyn(ArrayList architecture): Parameters: architecture: A list of Layer objects representing the architecture of the neural network. Methods void load_dataset(ArrayList<ArrayList> dataset):
Description: Loads the training dataset into the network. Parameters: dataset: A list of input-target pairs for training. double cost(JMatrix output, JMatrix target):
Description: Computes the cost (loss) between the predicted output and the target output. Parameters: output: The predicted output matrix. target: The actual target matrix. Returns: The computed cost. void applyGradient(double learnRate):
Description: Updates the network weights and biases using the calculated gradients and the learning rate. Parameters: learnRate: The learning rate used for updating weights and biases. ArrayList activations(JMatrix inputs):
Description: Computes the activations for each layer given the input matrix. Parameters: inputs: The input matrix to the network. Returns: A list of activation matrices for each layer. void train(int epochs, double learnRate, boolean print):
Description: Trains the neural network using the provided dataset for a specified number of epochs. Parameters: epochs: The number of training iterations. learnRate: The learning rate for gradient descent. print: A boolean indicating whether to print the cost after each epoch. JMatrix forward(JMatrix activations):
Description: Performs forward propagation through the network given the input activations. Parameters: activations: The input activations to the network. Returns: The output of the network after forward propagation. void init_weights():
Description: Initializes the weights and biases for the neural network randomly and sets up the gradient storage. void save(String path):
Description: Saves the neural network's weights and biases to a file. Parameters: path: The file path where the network data should be saved. void load(String path):
Description: Loads the neural network's weights and biases from a file. Parameters: path: The file path from which the network data should be loaded. Layer Class Documentation Overview The Layer class represents a single layer in a neural network. Each layer consists of a specified number of neurons and has an associated activation function along with its derivative.
Fields int size: The number of neurons in the layer. Function<JMatrix, JMatrix> activation: The activation function used in the layer. Function<JMatrix, JMatrix> derivative: The derivative of the activation function. Constructor Layer(int size, Function<JMatrix, JMatrix> activation, Function<JMatrix, JMatrix> derivative): Parameters: size: The number of neurons in the layer. Must be a positive integer. activation: A function representing the activation function for the layer. If null, a default "none" function is used, which simply returns the input. derivative: A function representing the derivative of the activation function. If null, a default "noneDerivative" function is used, which returns a zero matrix of the same dimensions as the input. Throws: IllegalArgumentException if size is less than or equal to zero. Methods JMatrix none(JMatrix x):
Description: Default activation function that returns the input matrix without any changes. Parameters: x: The input matrix to be processed. Returns: The same matrix x. JMatrix noneDerivative(JMatrix x):
Description: Default derivative function that returns a zero matrix with the same dimensions as the input matrix. Parameters: x: The input matrix. Returns: A zero matrix of the same dimensions as x. Function<JMatrix, JMatrix> getActivation():
Description: Returns the activation function of the layer. Returns: The activation function. Function<JMatrix, JMatrix> getDerivative():
Description: Returns the derivative of the activation function. Returns: The derivative function. String toString():
Description: Returns a string representation of the layer, including its size, activation function, and derivative. Returns: A string representation of the Layer object.