Evolutionary Synthesis of the Turing Machine's Rules

This doctoral thesis is concerned with possibilities of artificial intelligence utilization for Turing machine programming. The main topic regards using Differential Evolution and Self-Organizing Migrating Algorithm as selected methods of artificial intelligence for Turing machine transition function's rules synthesis. The rules of Turing machine represent a form of program on its basis Turing machine works. Rules designing can be considered as a way of this machine programming. The doctoral thesis consists of four parts which can be characterized as follows. The first part represents an introduction to the finite automata because Turing machines are classified as them. This part is necessary for understanding backgrounds of these machines. Highly important is a characterization of the machine on the basis of formal description. This is used as a base for rules synthesis problematics formulation in the next parts of the doctoral thesis. This first part also introduces selected algorithms of artificial intelligence. These are Differential Evolution and Self-Organizing Migrating Algorithm. The introduction to these algorithms is a key for settings of suitable parameters of selected algorithms while rules synthesis. The second part of the doctoral thesis presents two proposed approaches to Turing machine's rules synthesis (or optimization). These approaches are "classical optimization" and "per-partes optimization". Both approaches differ from each other. Each approach has also advantages and disadvantages which herewith assess their utilization. Both of mentioned approaches are closely described in this second part. In the third part of the doctoral thesis three selected elementary problems are introduced. These are unary addition, divisibility (exact divison) problem and primality (prime number detection) problematics. The problems are used as example tasks for Turing machine which rules we want to estimate by proposed approaches. It is utilized for analysis of rules optimization process dependence on custom settings of Differential Evolution and Self-Organizing Migrating Algorithm. This analysis is entirely fundamental not only for this part of the doctoral thesis but for the next part especially. The last, fourth, part of the doctoral thesis represents a practical utilization of proposed approaches to programming Turing machine by artificial intelligence. As real problematics protein processing by Turing machine was chosen. Proteins are regarded as primary protein structures in this case. Evolutionary synthesis of Turing machine's rules is demonstrated from total of a twelve selected primary protein structures differing in length. As described later in the text, this problematics is so far complex that can be considered as a sufficient way of proper work proof of proposed approaches to evolutionary synthesis of Turing machine's rules as main topic of the doctoral thesis.