Q-learning Lab

0.10
0.80
0.000
0.04
0

Definition: speedY = round(vy * 100), where vy is bird vertical speed in px/frame. So -140 means vy=-1.40 px/frame (up), +50 means vy=+0.50 px/frame (down).

Red means Q(Jump) is higher; Blue means Q(Stay) is higher. This is value preference (Q-value gap), not probability. X=tubeX, Y=diffY.

Episode Reward / Score

Gray=line of raw episode score. Red=line of moving average.