Modeling Place Cells with a Biological Plausible Artificial Neural Network

Abstract

The purpose of this paper is to use an artificial neural network to simulate the generation of place cell in the mammal hippocampus. These cells are a result of sensory stimuli input to animals like the rat. In this paper a model, using an artificial neural network, is created to simulate the rat hippocampus and the creation of place cells. It is made as biological plausible as possible. Place cells have been described in CA3 and CA1 for example, hence, the model will consist of four layers, namely the entorhinal cortex, dentate gyrus, CA3 and CA1. In explanations of the hippocampus the entorhinal cortex is often considered to be the input layer to the region and the model developed will follow this.Initial studies of relevant literature have been done as background for this model and a lot of inspiration have been collected from the model presented in Rolls (1995).The model developed here is able to reproduce place cells in both CA3 and CA1 where the latter has more defined place fields. In order to measure the quality of the combined place cells of a layer and compare the layers with each other a metric has been developed. This metric is used as a tool for developing the model parameters.In addition the results indicate that the model is able to make the CA1 neurons activate to a pattern that is a conjunction of smaller and less details patterns CA3 neurons are detectors of. Reproducing place cells means that the association between neurons first has to be learnt and then the model should be able to recall this information later.