population-irl
population-irl copied to clipboard
HumanCompatibleAI
(Experimental) Inverse reinforcement learning from trajectories generated by multiple agents with different (but correlated) rewards
It currently takes around 2 seconds to load the pirl module. Most of the time this isn't a big deal, but due to a combination of: 1. We restart each...
New requirements: 1. Make the lava not be giant columns 2. Create a parameter for different types of tiles with reward distributions (e.g. "Water" in the range "-1 to -5",...
Should be based on this: https://github.com/openai/gym/blob/master/gym/envs/mujoco/ant.py I think we probably just need to change the reward function part of that code. So either just copying that, or overwriting the step...
It's pretty slow right now, I suspect the value iteration is slow. Good to do some profiling to pin it down. Possible that moving things to GPU would speed things...
Metadata
Owner
Metadata
(Experimental) Inverse reinforcement learning from trajectories generated by multiple agents with different (but correlated) rewards