A Place Cell Recurrent Loop Learning Model Based on Non-negative Sparse Coding
To explore the neural mechanisms of brain navigation encoding,focus was placed on the neu-ral connections between the entorhinal cortex and the hippocampus for model research.Physiological evi-dence showed that significant feedback loop connections existed between the entorhinal cortex and the hippocampus,with the spatial encoding cells of both being highly correlated in navigational behavior.Based on this foundation,a feedback loop network model was established,where grid cells and weak spa-tial cells from the entorhinal cortex were taken as network inputs,connected to place cells and granule cells in the hippocampus,and non-negative sparse coding was employed for learning.Experimental re-sults indicated that the feedback learning model could rapidly capture the spatial tuning properties of these cells.Even when only weak spatial cells were used as inputs,the hippocampal place cells'unimod-al selectivity to space could be learned through the feedback loop,suggesting that the feedback encoding mechanism played a key role in optimizing spatial representation.In summary,the model might be one of the important cellular mechanisms for generating precise spatial encoding in the brain's navigation sys-tem.
万方数据
维普
