search

riiswa / kanrl

Kolmogorov-Arnold Network for Reinforcement Leaning, initial experiments
268 stars 33 forks source link

Comparison #1

Closed edmondja closed 6 months ago

edmondja commented 6 months ago

Thanks, I wanted to do the same with upside down RL. Can you please also show what performance you get with MLP so we can compare ?

riiswa commented 6 months ago

It's very interesting, if you want, and if you have time, you can add experiments with upside-down RL to the repository :D. I'm planning to compare my implementation with a classic MLP (probably this afternoon). With a hyperparameters search, as I have no idea what the right parameters are for KAN.

riiswa commented 6 months ago

@edmondja I updated the repo :D