I applaud this kind of effort. It is super useful to have context for the explosion of work in RL. One thing that is missing however is that reinforcement lear

I applaud this kind of effort. It is super useful to have context for the explosion of work in RL. One thing that is missing however is that reinforcement learning is a small fraction of what's know about learning through reinforcement. Decades of work in behavioral psychology and neuroscience lead to a very different understanding of learning where agents are not maximizing reward but vigorously pursuing homeostasis. Additionally the challenges map desperately needs a 'catastrophic forgetting' box. Perhaps this only applies to DRL but it is so fundamental IMO that to leave it out is to miss the foundation for the house.