Struct FeedForwardNet 

pub struct FeedForwardNet {
    pub layers: Vec<DenseLayer>,
}

A sequential feed-forward neural network.

Fields

layers: Vec<DenseLayer>

Ordered list of dense layers.

Implementations

impl FeedForwardNet

pub fn new() -> Self

Create an empty network.

pub fn add_layer(&mut self, layer: DenseLayer)

Append a layer to the end of the network.

pub fn forward(&self, input: &[f32]) -> Vec<f32>

Run a forward pass through all layers.

pub fn input_size(&self) -> Option<usize>

Returns the expected input width (None if the network has no layers).

pub fn output_size(&self) -> Option<usize>

Returns the output width of the last layer (None if empty).

pub fn total_parameters(&self) -> usize

Sum of parameters across all layers.

impl FeedForwardNet

pub fn compute_gradient_norm(&self, layer_grads: &[Vec<f32>]) -> f32

Compute the total gradient norm across all layers.

layer_grads[i] is the concatenated [grad_weights, grad_biases] for layer i. Returns the L2 norm of all gradients combined.

pub fn clip_gradients(&self, layer_grads: &mut [Vec<f32>], max_norm: f32) -> f32

Clip per-layer gradient vectors in-place so their combined norm ≤ max_norm. Returns the pre-clip norm.

Trait Implementations

impl Clone for FeedForwardNet

fn clone(&self) -> FeedForwardNet

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for FeedForwardNet

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

Formats the value using the given formatter.

impl Default for FeedForwardNet

fn default() -> Self

Returns the “default value” for a type.