Struct LinearLayer 

pub struct LinearLayer<T: Float + Debug + ScalarOperand + Send + Sync + 'static> { /* private fields */ }

Linear layer implementation

Implementations

impl<T: Float + Debug + ScalarOperand + Send + Sync + 'static> LinearLayer<T>

pub fn new(input_dim: usize, output_dim: usize) -> Result<Self>

Create new linear layer with Xavier/Glorot uniform weights, limit sqrt(6 / (fan_in + fan_out)).

pub fn new_he_init(input_dim: usize, output_dim: usize) -> Result<Self>

Create with He initialization (better for ReLU)

pub fn forward(&self, input: &Array2<T>) -> Result<Array2<T>>

Forward pass through linear layer: input @ weight + bias.

pub fn backward( &mut self, input: &Array2<T>, grad_output: &Array2<T>, learning_rate: T, ) -> Result<Array2<T>>

Backward pass. Returns dL/d(input) and applies an SGD step.

pub fn parameter_count(&self) -> usize

Get parameter count

pub fn reset(&mut self) -> Result<()>

Reset parameters, re-drawing the Xavier initialization.

pub fn get_weights(&self) -> &Array2<T>

Get weight matrix reference

pub fn get_bias(&self) -> &Array1<T>

Get bias vector reference

pub fn update_weights( &mut self, weight_delta: &Array2<T>, bias_delta: &Array1<T>, ) -> Result<()>

Update weights (for training)