Struct ActorNetwork 

pub struct ActorNetwork {
    pub hidden_weights: Array2<f64>,
    pub hidden_bias: Array1<f64>,
    pub output_weights: Array2<f64>,
    pub output_bias: Array1<f64>,
    pub _input_size: usize,
    pub hidden_size: usize,
    pub output_size: usize,
    pub activation: ActivationType,
}

Actor network for policy learning

Fields

hidden_weights: Array2<f64>

Hidden layer weights

hidden_bias: Array1<f64>

Hidden layer biases

output_weights: Array2<f64>

Output layer weights

output_bias: Array1<f64>

Output layer biases

_input_size: usize

Network architecture

hidden_size: usize

output_size: usize

activation: ActivationType

Activation function type

Implementations

impl ActorNetwork

pub fn new( input_size: usize, hidden_size: usize, output_size: usize, activation: ActivationType, ) -> Self

Create new actor network

pub fn forward( &self, input: &ArrayView1<'_, f64>, ) -> (Array1<f64>, Array1<f64>, Array1<f64>)

Forward pass through actor network

pub fn action_probabilities( &self, policy_output: &ArrayView1<'_, f64>, temperature: f64, ) -> Array1<f64>

Compute action probabilities with temperature scaling

pub fn backward_and_update( &mut self, input: &ArrayView1<'_, f64>, hidden_raw: &Array1<f64>, hidden_activated: &Array1<f64>, output_gradient: &ArrayView1<'_, f64>, learning_rate: f64, )

Backward pass and update weights

Trait Implementations

impl Clone for ActorNetwork

fn clone(&self) -> ActorNetwork

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ActorNetwork

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

Formats the value using the given formatter.