Struct renforce::trainer::DynaQ

pub struct DynaQ<S: Space, A: FiniteSpace, M: Model<S, A>> where S::Element: Hash + Eq,
        A::Element: Hash + Eq { /* fields omitted */ }

Represents an OnlineTrainer for Q-functions Uses the Dyna-Q algorithm

Methods

impl<S: Space, A: FiniteSpace, M: Model<S, A>> DynaQ<S, A, M> where S::Element: Hash + Eq,
        A::Element: Hash + Eq

fn new(action_space: A,
       gamma: f64,
       alpha: f64,
       train_period: TimePeriod,
       num_samples: usize,
       model: M)
       -> DynaQ<S, A, M>

Returns a new DynaQ with the given parameters

fn default(action_space: A, model: M) -> DynaQ<S, A, M>

Creates a new DynaQ with default parameters

fn gamma(self, gamma: f64) -> DynaQ<S, A, M>

Sets gamma field of self

fn alpha(self, alpha: f64) -> DynaQ<S, A, M>

Sets alpha field of self

fn train_period(self, train_period: TimePeriod) -> DynaQ<S, A, M>

Sets train_period field of self

fn num_samples(self, num_samples: usize) -> DynaQ<S, A, M>

Sets num_samples field of self

Trait Implementations

impl<S: Debug + Space, A: Debug + FiniteSpace, M: Debug + Model<S, A>> Debug for DynaQ<S, A, M> where S::Element: Hash + Eq,
        A::Element: Hash + Eq,
        A::Element: Debug,
        S::Element: Debug

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

Formats the value using the given formatter.

impl<T, S: Space, A: FiniteSpace, M: Model<S, A>> OnlineTrainer<S, A, T> for DynaQ<S, A, M> where T: QFunction<S, A> + Agent<S, A>,
        S::Element: Hash + Eq,
        A::Element: Hash + Eq

fn train_step(&mut self, agent: &mut T, transition: Transition<S, A>)

Performs one training iteration using the given transition

fn train(&mut self, agent: &mut T, env: &mut Environment<State=S, Action=A>)

Automatically trains the agent to perform well in the environment