rickstaa/stable-learning-control

LAC: Possible improvements

Closed this issue · 1 comments

rickstaa commented
  • Maybe the algorithm performance improves if we let the algorithm first do some random exploration actions.
  • Maybe the algorithm performance improves if we use different hyperparameters (nr of network layers, hidden layer size, activation functions)
rickstaa commented

Keep this in mind but not on our immediate roadmap.