Mini-Rainbow DQN

LIVE Side-by-side comparison — ALE/Breakout-v5
Speed:
DQN Standard Q-Network • 5M steps
DQN
0 Score
Episode
0
Avg (10 ep)
0
Best
0
Lives / Lost
5 / 0
Q-Values
NOOP
0
FIRE
0
RIGHT
0
LEFT
0
Action Distribution (this episode)
NOOP
0%
FIRE
0%
RIGHT
0%
LEFT
0%
CNN: 3-conv + 2-fc Replay: Uniform Target: argmax(Q_target)
Rainbow-Lite Dueling DDQN + PER • 2M steps
Rainbow-Lite
0 Score
Episode
0
Avg (10 ep)
0
Best
0
Lives / Lost
5 / 0
Q-Values
NOOP
0
FIRE
0
RIGHT
0
LEFT
0
Action Distribution (this episode)
NOOP
0%
FIRE
0%
RIGHT
0%
LEFT
0%
Dueling CNN: V(s) + A(s,a) Replay: PER (alpha=0.6) Target: argmax(Q_online)
Episode Log
# DQN Rainbow-Lite
ScoreLost ScoreLost
Waiting for first episode to complete...