Taxi-ai

AI for playing taxi-v3 (in gym env.)

Published On : https://sahilamin219.gitbook.io/ai/

Taxi-Agent

What is Q-Learning?

Q learning is an off-policy based Algorithim which uses valued based method for finding its optimal policy and uses TD-approach for training its action-value function. It is used for traing Q-function (an action value function) which bascially complete its Q-Table.

Minesweeper logic

power = +1

mine = -100

end = +100

NewQ(Start, right) = Q(start,  right) + aplha*(some delta value)
aplha=learning rate =0.9
some delta value = Reward at that state + (maxQ`(actions)) - Q(start, right);
actions = left, right,up, down
maxQ` = gamma*(max(actions)) 
gamma = discount rate;