Struct GpuNeuralSolver 

pub struct GpuNeuralSolver {
    pub layers: Vec<NeuralLayer>,
    pub learning_rate: f64,
}

Multi-layer perceptron running on the CPU mock GPU backend.

Layers are stacked so that the output of layer i feeds into layer i+1. All hidden activations use ReLU; the final layer is linear.

Fields

layers: Vec<NeuralLayer>

Ordered list of dense layers.

learning_rate: f64

Learning rate (stored for future back-prop use).

Implementations

impl GpuNeuralSolver

pub fn new(layer_sizes: &[usize], lr: f64) -> Self

Build a network from an ordered list of layer widths.

layer_sizes must contain at least two elements: the first is the input dimension and the last is the output dimension.

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

Run a full forward pass through all layers (ReLU hidden, linear output).

pub fn layer_count(&self) -> usize

Number of layers (= number of weight matrices).

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

Alias for forward_pass.

Trait Implementations

impl Clone for GpuNeuralSolver

fn clone(&self) -> GpuNeuralSolver

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GpuNeuralSolver

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

Formats the value using the given formatter.