A learning gap between neuroscience and reinforcement learning
This highlights a disconnect in AI research for neuroscientists and RL practitioners, though it is incremental as it adapts an existing task without proposing a new solution.
The paper identifies a gap between neuroscience and reinforcement learning by adapting a T-maze task from neuroscience, showing that state-of-the-art RL algorithms fail to solve it, and suggests neuroscience insights could address these issues.
Historically, artificial intelligence has drawn much inspiration from neuroscience to fuel advances in the field. However, current progress in reinforcement learning is largely focused on benchmark problems that fail to capture many of the aspects that are of interest in neuroscience today. We illustrate this point by extending a T-maze task from neuroscience for use with reinforcement learning algorithms, and show that state-of-the-art algorithms are not capable of solving this problem. Finally, we point out where insights from neuroscience could help explain some of the issues encountered.