Struct NeuralNetwork 

pub struct NeuralNetwork {
    pub learning_rate: f32,
    pub layer_sizes: Vec<u32>,
    pub weights: Vec<Matrix>,
    pub biases: Vec<Matrix>,
    pub command_receiver: Option<Receiver<NNCommand>>,
    pub command_sender: Option<Sender<NNCommand>>,
    pub update_interval: usize,
}

A Neural Network, use this one for training and the other one for predicting

Fields

learning_rate: f32

layer_sizes: Vec<u32>

weights: Vec<Matrix>

biases: Vec<Matrix>

command_receiver: Option<Receiver<NNCommand>>

command_sender: Option<Sender<NNCommand>>

update_interval: usize

Implementations

impl NeuralNetwork

pub fn new(layer_sizes: Vec<u32>) -> NeuralNetwork

pub fn train_batches( &mut self, input: Vec<Vec<f32>>, target: Vec<Vec<f32>>, batch_number: usize, ) -> usize

One Epoch: Create a batch of n neural networks, and divide the training data for those n networks Let them train and then average the weights and biases of the networks Return the averaged weights and biases and then average the weights and bias correction over all the networks and apply them to the neural network Run each batch in a thread Returns the amount of correct predictions

pub fn train_epochs( &mut self, input: Vec<Vec<f32>>, target: Vec<Vec<f32>>, batch_number: usize, epochs: usize, )

Train the neural network with the given input and target for x epochs

pub fn save_weights_biases(&self, file_name: String)

Save weights and biases to a file in the current directory called

pub fn load_from_save(path: &str) -> Self

Load weights and biases from a file in the current directory todo: tests this

pub fn train_epochs_m( &mut self, input_m: Vec<Matrix>, target_m: Vec<Matrix>, batch_number: usize, epochs: usize, )

Train the neural network with the given input and target for x epochs

pub fn predict(&self, input: Vec<f32>) -> Vec<f32>

Predict an output

Trait Implementations

impl Debug for NeuralNetwork

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.