Struct NeuralLayer 

pub struct NeuralLayer {
    pub weights: Vec<f64>,
    pub biases: Vec<f64>,
    pub n_in: usize,
    pub n_out: usize,
}

A single fully-connected (dense) neural network layer.

Weights are stored in row-major order: weights[i * n_in + j] is the weight from input j to output neuron i.

Fields

weights: Vec<f64>

Flattened weight matrix of shape [n_out × n_in].

biases: Vec<f64>

Bias vector of length n_out.

n_in: usize

Number of input features.

n_out: usize

Number of output neurons.

Implementations

impl NeuralLayer

pub fn new(n_in: usize, n_out: usize) -> Self

Create a new layer with all weights and biases initialised to 0.1.

pub fn forward(&self, input: &[f64]) -> Vec<f64>

Linear forward pass (no activation): W·x + b.

pub fn relu_forward(&self, input: &[f64]) -> Vec<f64>

Forward pass with ReLU activation applied element-wise.

pub fn tanh_forward(&self, input: &[f64]) -> Vec<f64>

Forward pass with tanh activation applied element-wise.

pub fn output_size(&self) -> usize

Number of output neurons.

pub fn input_size(&self) -> usize

Number of input features.

Trait Implementations

impl Clone for NeuralLayer

fn clone(&self) -> NeuralLayer

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for NeuralLayer

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

Formats the value using the given formatter.