Struct PAL 

pub struct PAL<Q> {
    pub q_func: Q,
    pub alpha: f64,
    pub gamma: f64,
}

Persistent Advantage Learning

References

• Bellemare, Marc G., et al. “Increasing the Action Gap: New Operators for Reinforcement Learning.” AAAI. 2016.

Fields

q_func: Q

alpha: f64

gamma: f64

Trait Implementations

impl<'m, S, Q> Handler<&'m Transition<S, usize>> for PAL<Q>

where Q: Enumerable<(&'m S,), Output = Vec<f64>> + Handler<StateActionUpdate<&'m S, usize, f64>>, <Q as Function<(&'m S,)>>::Output: Index<usize, Output = f64> + IntoIterator, <<Q as Function<(&'m S,)>>::Output as IntoIterator>::IntoIter: ExactSizeIterator,

type Response = <Q as Handler<StateActionUpdate<&'m S, usize>>>::Response

type Error = <Q as Handler<StateActionUpdate<&'m S, usize>>>::Error

fn handle( &mut self, t: &'m Transition<S, usize>, ) -> Result<Self::Response, Self::Error>

fn handle_unchecked(&mut self, msg: M) -> Self::Response

impl<Q> Parameterised for PAL<Q>

where Q: Parameterised,

fn weights(&self) -> Weights

Return an owned copy of the weights.

fn weights_dim(&self) -> (usize, usize)

Return the dimensions of the weight matrix.

fn weights_view(&self) -> WeightsView<'_>

Return a read-only view of the weights.

fn weights_view_mut(&mut self) -> WeightsViewMut<'_>

Return a mutable view of the weights.

fn n_weights(&self) -> usize