Abstract
A learning system based on the mathematical model of signaling network has been proposed. In that model, the time delay of information transmission in living cells is regarded as important element. It is known that a living body adapt to its environment, and it can be considered that each cell consisting the body has the same ability. From the viewpoint of mechanical engineering, it can be said that they have an autonomous system and must have sensor, actuator, and controller. However, the controller of cells to adapt the environment has not been understood well. Signaling network is expected to play the part of a control system in living cells, considering analogy with actual neural network. Cell signaling is a communication process using chemical reactions and diffusion of molecules. Various proteins constitute networks of chemical reactions. In this study, this signaling network has been modeled mathematically in reference to the artificial neural network, which has been applied to pattern recognition, control or various fields. Electric signals are used in an actual neural network, but diffusion of various molecules are used to transmit information in the signaling network. This big difference causes time delay to transmit information. In this paper, a mathematical model of signaling network including time delay has been proposed and a dynamic mapping system is constructed. Because of time delay, information which are inputted at different times are mixed thorough transmitting them. The model has been applied to map dynamic signals. As a result, it is shown that this model can learn sinusoidal waves and superimposed waves well.
