An adaptive evolutionary algorithm with intelligent mutation local searchers for designing multidrug therapies for HIV

Abstract

This paper proposes a novel Memetic Algorithm consisting of an Adaptive Evolutionary Algorithm (AEA) with three Intelligent Mutation Local Searchers (IMLSs) for designing optimal multidrug Structured Treatment Interruption (STI) therapies for Human Immunodeficiency Virus (HIV) infection. The AEA is an evolutionary algorithm with a dynamic parameter setting. The adaptive use of the local searchers helps the evolutionary process in the search of a global optimum. The adaptive rule is based on a phenotypical diversity measure of the population. The proposed algorithm has been tested for determining optimal 750-day pharmacological protocols for HIV patients. The pathogenesis of HIV is described by a system of differential equations including a model for an immune response. The multidrug therapies use reverse transcriptase inhibitor and protease inhibitor anti-HIV drugs. The medical protocol designed by the proposed algorithm leads to a strong immune response; the patient reaches a "healthy" state in one and half years and after this the STI medications can be discontinued. A comparison with a specific heuristic method and a standard Genetic Algorithm (GA) has been performed. Unlike the heuristic, the AEA with IMLSs does not impose any restrictions on the therapies in order to reduce the dimension of the problem. Unlike the GA, the AEA with IMLSs can overcome the problem of premature convergence to a suboptimal medical treatment. The results show that the therapies designed by the AEA lead to a "healthy" state faster than with the other methods. The statistical analysis confirms the computational effectiveness of the algorithm.

Agent Based Evolutionary Dynamic Optimization

“…We introduce a population index ξ [26] in order to measure its diversity and it can be calculated as follows:…”

Fuzzy predictive control based multiple models strategy for a tubular heat exchanger system

“…where A(q), B(q) and C(q) are given as the ARMAX model polynomial over q, y(k) is given as the output variable of the model, u(k) is given as the control variable, n(k) is given as the uncorrelated white noise with zero mean, e(k) is also given as the error modeling and finally (k) is given as the model parameters adaptation [34][35][36]. As it can be seen from the proposed strategy, the linear generalized predictive control; LGPC, scheme is used as controller of the system.…”

A memetic algorithm with adaptive hill climbing strategy for dynamic optimization problems

“…Some researchers [15,23] have suggested that multiple LS operators should be executed simultaneously on those individuals that are selected for local improvements and that a certain learning mechanism should be adopted to give the efficient LS methods greater chances to be chosen in the later stage. However, Neri et al [22] have also proposed a multiple LS based MA with a non-competitive scheme, where different LS methods can be activated during different population evolution periods. Inspired by these researches, an adaptive hill climbing (AHC) strategy that hybridizes the GCHC and SMHC methods described in Section 2.2 is proposed in this paper.…”