Struct renforce::trainer::LSPolicyIteration

pub struct LSPolicyIteration<F: Float> { /* fields omitted */ }

Least-squares Policy Iteration method

• Uses LSTD-Q for calculating the Q-function associated with a policy
• Only trains linear Q-functions (not currently enforced by library)

Methods

impl<F: Float> LSPolicyIteration<F>

fn new(gamma: F) -> LSPolicyIteration<F>

Constructs a new LSPolicyIteration with randomly initialized mean and deviation

fn gamma(self, gamma: F) -> LSPolicyIteration<F>

Updates gamma field of self

Trait Implementations

impl<F: Debug + Float> Debug for LSPolicyIteration<F>

fn fmt(&self, __arg_0: &mut Formatter) -> Result

Formats the value using the given formatter.

impl<F: Float + 'static, S: Space, A: Space, T> BatchTrainer<S, A, T> for LSPolicyIteration<F> where T: Agent<S, A> + ParameterizedFunc<F> + FeatureExtractor<S, A, F>

fn train(&mut self, agent: &mut T, transitions: Vec<Transition<S, A>>)

Trains agent based on the observed transitions

impl<F: Float> Default for LSPolicyIteration<F>

fn default() -> LSPolicyIteration<F>

Creates a new LSPolicyIteration with gamma = 0.99