Skip to content

Latest commit

 

History

History
 
 

dueling_ddqn

Dueling DDQN

Each experiment uses 3 seeds and is trained for 10k environment steps. The parameters used for Dueling DDQN are the same parameters as described in the original paper.

Pong Dueling DDQN - single worker

coach -p Atari_Dueling_DDQN -lvl pong

Pong Dueling DDQN

Breakout Dueling DDQN - single worker

coach -p Atari_Dueling_DDQN -lvl breakout

Breakout Dueling DDQN

Space Invaders Dueling DDQN - single worker

coach -p Atari_Dueling_DDQN -lvl space_invaders

Space Invaders Dueling DDQN