Topic: reinforcement learning agents

Learning and memory consolidation in the brain occur over multiple timescales. Inspired by this observation, it has been shown that catastrophic forgetting in reinforcement learning agents can be mitigated by consolidating Q-value function parameters at multiple timescales. In this work, we combine this approach with successor features, and show that by consolidating successor features and preferences learned over multiple timescales we can further mitigate catastrophic forgetting. In particular, we show that agents trained with this approach rapidly recall previously rewarding sites in large

Animals rely on sensory inputs to navigate complex, dynamic environments. Understanding how the brain integrates these inputs to guide behavior remains a central question in neuroscience. Training AI and animals on the same task offers a unique opportunity to uncover both similarities and critical differences in the strategies used by biological and artificial systems. By pinpointing what the brain accomplishes that modern AI lacks, we can drive advancements in machine learning, pushing AI to more robust, generalizable solutions. In this study, we train mice and reinforcement learning agents o