Reinforcement Learning

[fabioseel]

Stuff that needs to be done to make the Reinfocement Learning usable in the new system:

• develop example configs #36
• Discussion: Config of RL Model #71
  • also: automatic discovery / mapping of NeuralCircuits to Encoder/Core/Decoder ?
• how to implement RNN in Brain system (forward returning additional output RNN states, output shape return value?) NeuralCircuit Core, SFBrain - Init, SFBrain - Forward
• find way to instantiate brain outside Code #33
• update Observer for analysis and use new methods Code #35
• implement Multi-Objective / Goal system #34
• delete deprecated code (analysis, statistics, ...?)
• fix typing in RL codebase #38

Side-quests:

• Minimal samplefactory?

Nice-to-have Feature ideas:

• create a dataset like interface for doom scenarios (with test / train splits ...)