Struct NeuralController 

pub struct NeuralController {
    pub weights: Vec<Vec<Vec<f64>>>,
    pub biases: Vec<Vec<f64>>,
    pub lr: f64,
    pub replay: ReplayBuffer,
}

Simple multi-layer perceptron (MLP) neural network controller.

Fields

weights: Vec<Vec<Vec<f64>>>

Layer weights: each element is a weight matrix (out×in).

biases: Vec<Vec<f64>>

Layer biases: each element is a bias vector.

lr: f64

Learning rate.

replay: ReplayBuffer

Replay buffer.

Implementations

impl NeuralController

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

Create an MLP controller with given layer sizes. layer_sizes includes input and output dimensions.

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

Forward pass: returns output and all layer activations (for backprop).

pub fn gradient_update(&mut self, input: &[f64], target: &[f64])

Gradient update step (simple policy gradient / supervised). target is the desired output, input is the current state.

pub fn remember(&mut self, state: Vec<f64>, action: Vec<f64>, reward: f64)

Store experience in replay buffer.

pub fn compute(&self, state: &[f64]) -> Vec<f64>

Compute control output for given state.

Trait Implementations

impl Clone for NeuralController

fn clone(&self) -> NeuralController

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for NeuralController

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

Formats the value using the given formatter.