Skip to content

Add reinforcement learning

Nicola Gatto requested to merge add-reinforcement-learning into master
  • Update the version to 0.3.0-SNAPSHOT
  • Implement parameters for reinforcement learning (see for detailed parameter information)
  • Add more loss function parameters
  • Add a convenient way to use multi parameter entries (see MultiParamValueSymbol)
  • Add a RewardFunctionSymbol
  • Add CoCo:
    • From now on there are parameters exclusively for reinforcement learning and supervised learning. A check ensures that a parameter is used in the correct learning method context.
    • If target network is used, a training interval is required
    • Check that a reinforcement configuration always has information about the environment
    • Given the ROS environment, a check ensures that a reward function is given

Merge request reports