Integration and Transfer of Information: Search Engine Optimizer Inspired by Memes

The article presents research on the development of trainable optimization mechanisms. Such mechanisms are designed through the adaptive integration of memes that are accumulated by experience and/or through interaction with other mechanisms. The authors propose the concept of memetic automata. A memetic automaton is a software agent (or optimizer) that is capable of learning and autonomous behavior. Optimizers of various tasks can communicate with each other via the Internet with the support of cloud infrastructure. At the same time, the results of the solution for one task are transferred to the solution of another task. The article describes indicators of the relationship between tasks in terms of the expected utility of transferring memes. The authors introduce the concept of unification of the search space to ensure the generality of the formulation of a set of optimization problems. Also, we offer procedures for serial and parallel transfer of knowledge memes when optimizing the target task. The effectiveness of optimizers is experimentally confirmed by the example of designing a neuroevolutionary controller for a cart robot.