Modern Immunological Models and Their Applications

Abstract

he paper considers main models and algorithms of artificial immune systems, which are related to the evolutionary computation paradigm and used to search for potential solutions, each of which is represented by an artificial lymphocyte. Same as an individual in evolutionary computation, an artificial lymphocyte is most often encoded by a binary string or a vector of real numbers. As far as the main models of artificial immune systems are concerned, the clonal selection algorithm is close to the evolutionary strategy of evolutionary computing, though it uses more powerful mutation operators and is applied mainly to solve numerical and combinatorial optimisation problems. The negative selection algorithm is based on the "friend or foe" recognition principle found in the immune system and is most popular in applications. The paper presents two aspects of the algorithm: 1) the basic concept, that is, expanding the set of "friend" cells; 2) the goal, which is to learn to distinguish between "friend" and "foe" cells, while only "friend" cell samples are available. We consider continuous and discrete network models representing regulated networks of molecules and cells. We note the advantages and disadvantages of these models and their application in the field of computer security, robotics, fraud and malfunction detection, data mining, text analysis, image recognition, bioinformatics, games, planning, etc.