An inhibitory plasticity mechanism for world structure inference by hippocampal replay

• DOI: 10.1101/2022.11.02.514897
• Publication Date: 2022-11-04T03:15:15Z
• Authors: Zhenrui Liao; Darian Hadjiabadi; Satoshi Terada; Ivan Soltesz; Attila Losonczy
• Keywords: Hebbian theory; Learning rule

Summary Memory consolidation assimilates recent experiences into long-term memory. This process requires the replay of learned sequences, though content these sequences remains controversial. Recent work has shown that statistics deviate from those experience: stimuli which are experientially salient may be either selected or suppressed. We find this phenomenon can explained parsimoniously and biologically plausibly by a Hebbian spike time-dependent plasticity rule at inhibitory synapses. Using spiking networks three levels abstraction–leaky integrate-and-fire, biophysically detailed, abstract binary–we show enables efficient inference model structure world. present analytical results replayed converge to ground truth under mathematical replay. Finally, we make specific predictions about consequences intact perturbed dynamics for network cognition. Our outlines potential direct link between synaptic cognitive memory consolidation, with implications both normal learning neurological disease.