Embedding global physiological data in an artificial neural network

The goal of creating a realistic neural network model for studying biological processing has constantly piqued the interest of many researchers (Grossberg, 1995). One reason for the lack of realism in many existing artificial neural network models is that the dynamics of the human brain have not been captured. The difficulty partially arises from the lack of mathematical representation that can more easily express the dynamics of the brain. This paper introduces a network model that can be used to incorporate the global dynamic data of the brain. In particular, this project has incorporated into a Fukushima (1984) based network a set of data that has been obtained from a true recording of electroencephalographic (EEG) potential of the human subject, who was in a light sleep condition. The incorporation of the EEG data into an artificial neural network is important, since biological neurons never achieve a stable state, unless they are dead or in deep anesthesia. In addition, some physiological findings have suggested that neurons in different parts of the brain might communicate with each other through oscillation of synchronous electrical activity at 40 hertz (Barinaga, 1990). This study has shown that the new network model is able to achieve the pattern associative capabilities of the original model.