Struct DenseLayer64 

pub struct DenseLayer64 {
    pub weights: Vec<f64>,
    pub biases: Vec<f64>,
    pub in_features: usize,
    pub out_features: usize,
    pub activation: ExtActivation,
    pub last_pre_act: Vec<f64>,
    pub last_output: Vec<f64>,
    pub last_input: Vec<f64>,
}

A fully-connected layer with f64 weights supporting forward pass and gradient computation for backpropagation.

Fields

weights: Vec<f64>

Weight matrix in row-major layout: weights[out * in_features + in].

biases: Vec<f64>

Bias vector of length out_features.

in_features: usize

Number of input features.

out_features: usize

Number of output features.

activation: ExtActivation

Activation function.

last_pre_act: Vec<f64>

Pre-activation outputs from the last forward pass (z = W*x + b).

last_output: Vec<f64>

Post-activation outputs from the last forward pass.

last_input: Vec<f64>

Last input fed to this layer.

Implementations

impl DenseLayer64

pub fn new( in_features: usize, out_features: usize, activation: ExtActivation, ) -> Self

Create a new layer with zero-initialised weights and biases.

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

Forward pass: computes activation(W * input + b). Caches pre_act, output, and input for backprop.

pub fn backward(&self, delta_out: &[f64]) -> (Vec<f64>, Vec<f64>, Vec<f64>)

Backward pass: computes gradients w.r.t. weights, biases, and input.

delta_out is the gradient of the loss w.r.t. this layer’s output (same shape as last_output).

Returns (grad_weights, grad_biases, delta_in) where delta_in is the gradient passed to the previous layer.

pub fn apply_sgd(&mut self, grad_weights: &[f64], grad_biases: &[f64], lr: f64)

Apply gradient updates using a simple SGD step.

pub fn num_params(&self) -> usize

Total number of parameters.

Trait Implementations

impl Clone for DenseLayer64

fn clone(&self) -> DenseLayer64

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DenseLayer64

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

Formats the value using the given formatter.