Interleaved Replay of Novel and Familiar Memory Traces During Slow-Wave Sleep Prevents Catastrophic Forgetting
Golden, R.; Saxena, R.; Gonzalez, O. C.; Delanois, J. E.; Kilianski, S.; McNaughton, B. L.; Bazhenov, M.
Show abstract
Humans and animals can learn continuously, acquiring new knowledge while integrating it into a lifelong memory pool. In contrast, artificial neural networks (ANNs) suffer from catastrophic forgetting, where new training disrupts existing memories. This issue can be alleviated in ANNs by interleaving training on new tasks with past data; however, whether the brain uses a similar strategy is unknown. In this work, we show that slow-wave sleep interleaves replay of familiar and novel (i.e. hippocampal-dependent) memory traces within individual slow waves, allowing new memories to integrate into the existing cortical pool without interference. This study presents a novel theory for how memory traces acquired across an animals life are organized within the cortical-hippocampal system to support continual learning and suggests novel principles for a broad range of continual learning AI.
Matching journals
The top 3 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.