gsnn.optim.RewardScaler

Functions

edw(N, alpha)

Classes

RewardScaler([clip, eps, alpha, warmup])

class gsnn.optim.RewardScaler.RewardScaler(clip=5, eps=0.001, alpha=0.04, warmup=3)

Bases: object

get_params()

scale(reward)

update(reward)

gsnn.optim.RewardScaler.edw(N, alpha)