Daejoong Kim

You can see hyperparameters in the file below. https://github.com/only4hj/DeepRL/blob/master/env/torcs/torcs_env.py (See parameters with option args.drl == 'dqn') Main algorithm is DDPG. The train is done under 1,000 episodes (4 hours in...

I have not tried using a3c for torcs. I used only DDPG for torcs.

Yes I did. It works fine with DDPG using only image input. With low features input the network outputs three actions, steering, acceleration and brake. With only image input i...

For now there is no plan to publish trained models. You can install and train them by following the setup below. https://github.com/only4hj/DeepRL/blob/master/env/torcs/INSTALL.md

I got a similar score to the DeepMind paper in kung_fu_master after taking loss of life as the terminal state. Previously, with the original simple_dqn source, kung_fu_master showed much lower...

Two more things to get close to the DeepMind paper. 1) DeepMind code copies the trained net to the target net once every 10,000 steps. DeepMind uses 10,000 steps for...

@mthrok 1) 10,000 env steps means 10,000 observation from agent's view point in skip_frame=4 (which means 40,000 ale frames). 2) I think it's torch's default initialization. See these. https://groups.google.com/d/msg/deep-q-learning/JV384mQcylo/De1Jzc0hAAAJ https://github.com/gtoubassi/dqn-atari/blob/master/dqn.py#L148...