Hands-On Reinforcement Learning for Games

Incremental means

An incremental or running mean allows us to keep an average for a list of numbers without having to remember the list. This, of course, has huge benefits when we need to keep a mean over 50,000, 1 million, or more episodes. Instead of updating the mean from a full list, for every episode, we hold one value that we incrementally update using the following equation:

In the preceding equation, we have the following:

• = The current state value for the policy
• = Represents a discount rate
• = The current total return

By applying ...