Online Course Support | Practical Reinforcement Learning

What does the reward discounting means for an agent?

Intuitively, think of two extreme cases: \gamma=1 (the return is an infinite sum of random variables) and \gamma=0 (the return is a single random variable).

Given \gamma < 1, for close rewards the discounting will be \gamma^{\text{small number}}, which is larger than the discounting for distant rewards, \gamma^{\text{large number}}.

Similar Posts