Rapid context inference in a thalamocortical model using recurrent neural networks

This paper presents a computational model of the prefrontal cortex (PFC) and mediodorsal thalamus (MD) interaction that utilizes a Hebbian plasticity rule to support rapid, online context inference. The model demonstrates that the MD thalamus can infer temporal contexts from PFC inputs within a few trials, achieved through PFC-to-MD synaptic plasticity with pre-synaptic traces and adaptive thresholding, along with winner-take-all normalization in the MD. The model also shows how the thalamus gates context-irrelevant neurons in the PFC, facilitating continual learning and alleviating catastrophic forgetting. The model's performance is evaluated on sequential cognitive tasks, showcasing its ability to switch between contexts and transfer knowledge.

Wei-Long Zheng, Zhongxuan Wu, Ali Hummos, Guangyu Robert Yang, Michael M. Halassa