jaclearn.rl.algo.math

Classes

LinearValueRegressor
ObservationNormalizer

Functions

compute_gae(rewards, values, next_val, ...)
discount_cumsum(x, gamma)	Compute the discounted cumulative summation of an 1-d array.
discount_return(x, discount)	Compute the discounted return summation of an 1-d array.
normalize_advantage(adv)

Class LinearValueRegressor

class LinearValueRegressor

Bases: object

__init__()

__new__(**kwargs)

fit(states, steps, returns)

predict(states, steps)

register_snapshot_parts(env)

coeffs = None

Class ObservationNormalizer

class ObservationNormalizer

Bases: object

__call__(o)

Call self as a function.

__init__(filter_mean=True)

__new__(**kwargs)

normalize(o)

Functions

compute_gae(rewards, values, next_val, gamma, lambda_)

discount_cumsum(x, gamma)

Compute the discounted cumulative summation of an 1-d array. From rll/rllab

discount_return(x, discount)

Compute the discounted return summation of an 1-d array. From rll/rllab

normalize_advantage(adv)