计算模型提出 REM 样睡眠通过再激活靶向修剪减少记忆干扰

The role of Rapid Eye Movement (REM) sleep in memory consolidation remains poorly understood. Several influential ideas propose that REM sleep serves to eliminate crosstalk between memories, thereby orthogonalizing representations of similar ones; however, mechanistic confirmation remains sparse. Here we used recurrent neural networks trained with Equilibrium Propagation (EP), a biologically plausible training algorithm, to investigate the role of REM-like sleep in memory consolidation. We found that REM-like sleep reduces memory interference by decorrelating task-specific neuronal populations, isolating individual memory traces. During sleep, reactivation was dominated by the population encoding the most recent memory. Net excitatory input onto this population decreased, most strongly that from older-memory populations, which allowed inputs of older memories to drive their own neurons again. Spreading the same total synaptic change evenly across all target populations abolished this recovery. We call this mechanism reactivation-targeted pruning: plasticity is confined to one population, whose neurons lose inputs that are silent when they fire. The mechanism held under Dale's law, and depression of excitatory synapses alone, without any inhibitory plasticity, was sufficient. These findings suggest mechanistic principles underlying REM sleep's role in memory consolidation and highlight the potential for integrating human-like learning behaviors into artificial neural networks.

中文翻译还没有生成，先显示原文。