How do you update the Q values for a two player game

Question

How do you update the Q values for a two player game

For a single player game, the Q-value updates are pretty intuitive. The current state and future state depend on the strategy of one player, but for two players this is not the case. Consider a scenario in which your opponent wins and the game ends. How are Q values updated?

+3

reinforcement-learning q-learning multiplayer

Dipti chaudhari Apr 07 17 at 22:11

source to share

1 answer

Nick walker · Answer 1 · 2017-04-11T16:05:17+0000

One general approach is to view your opponent as part of the environment, so the state will be defined to include the statement, the position of the opponent. You choose an action and perform it, changing the state. Then the opponent takes his action, changing the state again. Then your agent gets an initial state, which is the result of the previous action and the previous action of the enemy .

So, in the case where in the state s

you perform an action a

, then the opponent acts and ends the game, you must record the transition from s

to the terminal state through a

.

How do you update the Q values ​​for a two player game

More articles:

How do you update the Q values for a two player game