Function optimization algorithm based on SIRQV epidemic dynamic model

To solve some complicated function optimization problems, the SIRQV algorithm is constructed based on the SIRQV epidemic model. The algorithm supposes that some animal individuals exist in an ecosystem; each individual is characterized by a number of features; an infectious disease exists in the ecosystem and spreads among individuals, the disease attacks a part of features of an individual at each time. Each infected individual may pass through such states as susceptibility (S), infection (I), recovery (R), quarantine (Q) and vaccination (V), which can synthetically decide the physique strength of an individual. Individuals in the algorithm have 5 states such as S, I, R, Q and V, and 13 state transitions, each of which is equivalent to an operator. the 13 operators are logically organized together by the disease transmission logic of the SIRQV epidemic model so as to form a good cooperation and sufficient information exchange among individuals. The algorithm uses the activation, average, combination, reinforcement and assimilation operator to exchange feature information among individuals. The reinforcement operator transfers feature information from some strong individuals with higher individual physique index (IPI) to a weak individual with lower IPI index so as to make the latter grow better; the average operator ensures an individual to obtain average feature information from other individuals so as to reduce the probability that the individual drops into local optima; the activation operator expands an individual's search scope by increasing its vitality; the combination operator has the characteristics of both the activation operator and the average operator; the assimilation operator enables the search to possess of jumping ability along dimension direction; the REINIT operator has exploration and exploitation ability to overcome sticky state of individuals and enhance precision of global optima; the growth operator enables the algorithm to converge globally. Results show that the algorithm has characteristics of strong search capability and global convergence, and has a high convergence speed for some complicated function optimization problems, especially for some function optimization problems with high condition number.