Simulation of hippocampal place cell phenomenon using multilayer perceptrons

The hippocampal place cell phenomenon, where certain hippocampal cells of a freely moving rat will respond preferentially to a location in its already familiar environment, was simulated using a multilayer perception neural network. The input to the network was the position of the rate represented as a vector of distances to the cues. Discretizing the experimental surface, a data set was created to train a network with one hidden layer. The performance of the network and its ability to generalize were tested and the surface of mean absolute error plotted. The plot showed interesting properties such as good ability to generalize in certain locations and ambiguity due to symmetry of cue positions.<
>