RL.DP

Synopsis

• type Probability = Rational
• type P s a = HashMap s (Set (a, Probability))
• type V s num = HashMap s num
• diffV :: (Eq s, Hashable s, Num num) => V s num -> V s num -> num
• class (Ord s, Ord a, Fractional num, Ord num, Hashable s) => DP_Problem pr s a num | pr -> s, pr -> a, pr -> num where
• action :: DP_Problem pr s a num => pr -> P s a -> s -> Set (a, Probability)
• initV :: DP_Problem pr s a num => pr -> num -> V s num
• invariant_probable_actions :: (DP_Problem pr s a num, Show s, Show a) => pr -> Bool
• invariant_closed_transition :: (DP_Problem pr s a num, Show s, Show a) => pr -> Bool
• invariant_no_dead_states :: (DP_Problem pr s a num, Show s, Show a) => pr -> Bool
• invariant_terminal :: (DP_Problem pr s a num, Show s, Show a) => pr -> Bool
• invariant_policy_actions :: (DP_Problem pr s a num, Ord a, Show s, Show a) => P s a -> pr -> Bool
• invariant_policy_prob :: (DP_Problem pr s a num, Ord a, Show s, Show a) => P s a -> pr -> Bool
• invariant :: (DP_Problem pr s a num, Show s, Show a, Ord a) => pr -> Bool
• policy_eq :: (Eq a, DP_Problem pr s a num) => pr -> P s a -> P s a -> Bool
• uniformPolicy :: (Ord a, DP_Problem pr s a num) => pr -> P s a
• data Opts num s a = Opts {
  • eo_gamma :: num
  • eo_etha :: num
  • eo_max_iter :: Int
  }
• defaultOpts :: Fractional num => Opts num s a
• data EvalState num s = EvalState {
  • _es_delta :: num
  • _es_v :: V s num
  • _es_v' :: V s num
  • _es_iter :: Int
  }
• es_v' :: forall num s. Lens' (EvalState num s) (V s num)
• es_v :: forall num s. Lens' (EvalState num s) (V s num)
• es_iter :: forall num s. Lens' (EvalState num s) Int
• es_delta :: forall num s. Lens' (EvalState num s) num
• initEvalState :: Fractional num => V s num -> EvalState num s
• policy_eval :: (Monad m, DP_Problem pr s a num) => Opts num s a -> P s a -> V s num -> DP pr m s a num -> m (V s num)
• policy_action_value :: DP_Problem pr s a num => Opts num s a -> s -> a -> V s num -> pr -> num
• policy_improve :: (Monad m, DP_Problem pr s a num) => Opts num s a -> V s num -> DP pr m s a num -> m (P s a)
• data DP pr m s a num = DP {
  • dp_pr :: pr
  • dp_trace :: V s num -> P s a -> m ()
  }
• policy_iteration :: (Monad m, DP_Problem pr s a num, Ord a) => Opts num s a -> P s a -> V s num -> DP pr m s a num -> m (V s num, P s a)

Documentation