Organisms need to properly assess the rewards and efforts they receive when learning appropriate behavioral strategies.For this purpose, an information processing mechanism of the brain that prioritizes and learns and memorizes things is required.This also applies to machine learning, and research on computational algorithms that improve learning efficiency has been active.In this regard, it is important to efficiently reproduce (replay) the information learned by the computational circuit, but it was difficult to evaluate changes in behavior patterns and neural activity in animals because the behavior according to learning is complicated. ..

The research group focused on the hippocampus as a brain region that realizes such information processing.We investigated changes in hippocampal neuronal population activity and information regeneration as rats were asked to solve maze tasks and learn new behavioral strategies for rewards.As a result of the analysis, it was discovered that the hippocampal neural circuit forms an efficient information expression as learning progresses, and frequently replays episodes important for learning.Moreover, even before the rat adopted the most efficient behavioral strategy, the replay of that behavioral strategy began in the brain.In other words, the hippocampal neural circuit proved to have the property of being able to prioritize and replay learned events and behavioral strategies.

The results of this research will help elucidate the brain information processing mechanism for animals to design behaviors suitable for the new environment.Furthermore, in addition to brain research, it will be an important step in devising computational algorithms for efficiently advancing machine learning.

Paper information:[Proceedings of the National Academy of Sciences of the United States of America] Prioritized experience replays on a hippocampal predictive map for learning