[q_learning]Number Game

Number game

I’ve set values like below.

EPISODES = 15000        MOVE_PENALTY = -1        ANSWER_REWARD = 50      HALF_REWARD = 20
epsilon = 0.9           EPS_DECAY = 0.99998      SHOW_EVERY = 500         STEP = 200
LEARNING_RATE = 0.001   DISCOUNT = 0.95         

It basically guess the numbers(2 digits) that are randomly set between 0 to 9.

The q table is 2 dimension which is included tuple number(observation) and each number of actions’ values.

There are 5 actions that are ‘minus one’, ‘plus one’ and ‘stay’ each digits except ‘stay’.

(average of episode rewards)

I have learnt q learning from sentex Youtuber.