T-maze training of a recurrent CA3 model reveals the necessity of novelty-based modulation of LTP in hippocampal region CA3

摘要: The rat hippocampus has been shown to mediate a large set of spatial navigation tasks such as the simple T-maze. We investigated performance minimal computational, but biologically based, model CA3 on this task. For successful performance, needs generate and maintain neuronal codes for each two arms Moreover, code must be distinctively recalled in goal-dependent manner. development is aided by appearance repetitively firing recurrent neurons - known local context units, analogous hippocampal place cells which promote spatiotemporal association within T-maze training sequences. number, longevity, connectivity units exclusively coding arm maze grow with training. Although too much training, one uncontrollably dominates over other code, goal appropriate choice-behavior lost. That is, network can easily deteriorate overtraining. amount that produces deterioration depends parameters. Rather than failure model, we believe these results tell us something important about biology system. result provides support hypothesis afferent system down-regulates LTP once task successfully learned. Modulatory Systems (e.g., dopaminergic, generally Dl/D5r) exist are candidates functional role.