Evolvable biological interfaces: An outline of the new computing system

Abstract

In this paper, we represent architecture of a new individual-based model of evolvable biological interfaces. The model is based on three parts: mutable coding sequence, individual agents produced by the coding sequence and external environment represented by various virtual substances. Agents are able to recognize external substances and to perform transformations on them. However, unlike general computational scheme currently used in evolutionary computation which is based on predefined list of actions, where each action have predefined fitness value defined strictly within current problem, the recognition and corresponding transformation in this model are organized in completely different way. Here we merged input-action-output system into one unified interdependent whole where each element's role is relative to its surrounding, and instead of unambiguous symbols we introduced alphabet where each symbol is characterized by a set of characteristics. As a result, we obtained system which is able to eventually, through evolution, recognize and assimilate unlimited number of new substances and to perform on them non-fixed set of transformations.