Feature/pgmorl - Githubissues

LucasAlegre / morl-baselines

Multi-Objective Reinforcement Learning algorithms implementations.

https://lucasalegre.github.io/morl-baselines

MIT License

296 stars 47 forks source link

Closed ffelten closed 2 years ago

ffelten commented 2 years ago

TODO:

[ ] Building a continuous Pareto front from the learned policies by interpolation
[x] Fix the number of iterations (something wrong)
[x] Fix the weight-task allocation algorithm
[x] Find out why the policies are not updated after changing weights (deep copy etc.) -> KL divergence is 0 or very close to it.

ffelten commented 2 years ago