Struct LinearLayer 

pub struct LinearLayer<const I: usize, const O: usize> { /* private fields */ }

A linear layer in a neural network, with sigmoid activation.

Implementations

impl<const I: usize, const O: usize> LinearLayer<I, O>

pub fn new(learning_rate: f32) -> Self

Creates a new linear layer with random weights and biases.

pub fn feedforward(&self, input: &RowVector<I>) -> (RowVector<O>, RowVector<O>)

Performs a forward pass through the layer, returning the output and activated output.

pub fn backpropagate( &mut self, input: &RowVector<I>, output: &RowVector<O>, error_next: &RowVector<O>, ) -> RowVector<I>

Performs backpropagation for this layer, updating weights and biases, and returning the error to propagate to the previous layer.

Arguments

• input: The input to this layer (activated output from previous layer, or raw input for first layer)
• output: The pre-activation output from the forward pass
• error_next: The error propagated from the next layer (or output error for final layer)
• learning_rate: The learning rate for weight updates

Returns

The error to propagate to the previous layer