| Generally, from point of view of the architecture of the current information processing models composed of many simple computational elements (called units as follows), two types of these models have been developed. One type is that the models consist of many units connected by physical links with dense connections, representatively such as artificial neural networks; and the other is that the models consist of many discrete units which have neither physical links nor connections between each other, and realize the information processing function via wave motion in the potential field, representatively such as the vibrating potential field model. It is known that the architecture for information processing in nature including human society usually acts as a hybrid architecture from these two types models. Based on this understanding, we aim to find out a universal principle of information processing and develop a new computational theory. Therefore, the study is to attempt to theoretically construct a hybrid framework called artificial neural electromagnetic field absorbing the quintessences from the network and field models, and verify its appropriateness as an information processing mechanism. As a basic research we propose a unit hybrid artificial neuron model by introducing magnetic field information into the conventional neuron model (McCulloch-Pitts' neuron model). With the new neuron model, we construct a hybrid architecture called an artificial neural electromagnetic field model from the network and field models. |
