Dušan Tošić

Solving the simple plant location problem by genetic algorithm

The simple plant location problem (SPLP) is considered and a genetic algorithm is proposed to solve this problem. By using the developed algorithm it is possible to solve SPLP with more than 1000 facility sites and customers. Computational results are presented and compared to dual based algorithms.

Two genetic algorithms for solving the uncapacitated single allocation p-hub median problem

This paper deals with the Uncapacitated Single Allocation p-Hub Median Problem (USApHMP). Two genetic algorithm (GA) approaches are proposed for solving this NP-hard problem. New encoding schemes are implemented with appropriate objective functions. Both approaches keep the feasibility of individuals by using specific representation and modified genetic operators. The numerical experiments were carried out on the standard ORLIB hub data set. Both methods proved to be robust and efficient in solving USApHMP with up to 200 nodes and 20 hubs. The second GA approach achieves all previously known optimal solutions and achieves the best-known solutions on large-scale instances.

A genetic algorithm for the index selection problem

This paper considers the problem of minimizing the response time for a given database workload by a proper choice of indexes. This problem is NP-hard and known in the literature as the Index Selection Problem (ISP). We propose a genetic algorithm (GA) for solving the ISP. Computational results of the GA on standard ISP instances are compared to branch-and-cut method and its initialisation heuristics and two state of the art MIP solvers: CPLEX and OSL. These results indicate good performance, reliability and efficiency of the proposed approach.

An evolutionary-based approach for solving a capacitated hub location problem

This paper addresses the capacitated hub location problem (CHLP), which is a variant of the classical capacitated hub problem. What is presented is a modified mixed integer linear programming (MILP) formulation for the CHLP. This modified formulation includes fewer variables and constraints compared to the existing problem formulations in the literature. We propose two evolutionary algorithms (EAs) that use binary encoding and standard genetic operators adapted to the problem. The overall performance of both EA implementations is improved by a caching technique. In order to solve large-scale instances within reasonable time, the second EA also uses a newly designed heuristic to approximate the objective function value. The presented computational study indicates that the first EA reaches optimal solutions for all smaller and medium-size problem instances. The second EA obtains high-quality solutions for larger problem dimensions and provides solutions for large-scale instances that have not been addressed in the literature so far.

Novel Phylogenetic Algorithm to Monitor Human Tropism in Egyptian H5N1-HPAIV Reveals Evolution toward Efficient Human-to-Human Transmission

Years of endemic infections with highly pathogenic avian influenza (HPAI) A subtype H5N1 virus in poultry and high numbers of infections in humans provide ample opportunity in Egypt for H5N1-HPAIV to develop pandemic potential. In an effort to better understand the viral determinants that facilitate human infections of the Egyptian H5N1-HPAIVvirus, we developed a new phylogenetic algorithm based on a new distance measure derived from the informational spectrum method (ISM). This new approach, which describes functional aspects of the evolution of the hemagglutinin subunit 1 (HA1), revealed a growing group G2 of H5N1-HPAIV in Egypt after 2009 that acquired new informational spectrum (IS) properties suggestive of an increased human tropism and pandemic potential. While in 2006 all viruses in Egypt belonged to the G1 group, by 2011 these viruses were virtually replaced by G2 viruses. All of the G2 viruses displayed four characteristic mutations (D43N, S120(D,N), (S,L)129Δ and I151T), three of which were previously reported to increase binding to the human receptor. Already in 2006–2008 G2 viruses were significantly (p<0.02) more often found in humans than expected from their overall prevalence and this further increased in 2009–2011 (p<0.007). Our approach also identified viruses that acquired additional mutations that we predict to further enhance their human tropism. The extensive evolution of Egyptian H5N1-HPAIV towards a preferential human tropism underlines an urgent need to closely monitor these viruses with respect to molecular determinants of virulence.

GA Inspired Heuristic for Uncapacitated Single Allocation Hub Location Problem

In this article, the results achieved by applying GA-inspired heuristic on Uncapacitated Single Allocation Hub Location Problem (USAHLP) are discussed. Encoding scheme with two parts is implemented, with appropriate objective functions and modified genetic operators. The article presents several computational tests which have been conducted with ORLIB instances. Procedures described in related work round distance matrix elements to few digits, so rounding error is significant. Due to this fact, we developed exact total enumeration method for solving subproblem with fixed hubs, named Hub Median Single Allocation Problem (HMSAP). Computational tests demonstrate that GA-inspired heuristic reach all best solutions for USAHLP that are previously obtained and verified branch-and-bound method for HMSAP. Proposed heuristic successfully solved some instances that were unsolved before.

Electromagnetism-like algorithm for support vector machine parameter tuning

This paper introduces an electromagnetism-like (EM) approach for solving the problem of parameter tuning in the support vector machine (SVM). The proposed method is used to tune binary SVM classifiers in single and multiple kernel mode. The internal kernel structure is based on linear and radial basis functions (RBF). An appropriate encoding scheme of EM enables easy transformation of real-valued EM points directly to real-valued parameter combinations. Estimations of the generalization error based on the cross-validation and validation set error are used as objective functions. The efficient local search procedure uses variable size interval movement in order to improve the convergence of the method. The quality of the proposed method is tested on four collections of testing benchmarks through five separate experiments. The first three collections consist of small-size to medium-size classification data sets with up to 60 features and 1,300 training vectors, while the fourth collection is formed of large heterogeneous data sets with up to 1,554 features and 2,186 training vectors. The obtained results indicate that EM outperforms the comparison algorithms in 10 out of 13 instances from the first collection, 5 out of 5 instances from the second, and 13 out of 15 instances from the third collection. The last two experiments, conducted on the fourth collection, show that the proposed method outperforms 14 successful methods in 3 out of 5 data sets where RBF multiple kernel learning is used, and behaves competitively in cases when linear kernels are used.

Two Hybrid Genetic Algorithms for Solving the Super-Peer Selection Problem

The problem that we will address here is the Super-Peer Selection Problem (SPSP). Two hybrid genetic algorithm (HGA) approaches are proposed for solving this NP-hard problem. The new encoding schemes are implemented with appropriate objective functions. Both approaches keep the feasibility of individuals by using specific representation and modified genetic operators. The numerical experiments were carried out on the standard data set known from the literature. The results of this test show that in 6 out of 12 cases HGAs outreached best known solutions so far, and that our methods are competitive with other heuristics.

A Genetic Algorithm for the Uncapacitated Network Design Problem

In this paper a genetic algorithm (GA) for solving the uncapacitated network design problem (UNDP) is presented. The problem with single source and destinations for each commodity is considered. UNDP is a base in class of the network design problems, but it is still NP-hard. The implementation of GA is additionally improved by caching technique of GA. The computational results on instances up to 50 commodities, 100 nodes and 700 edges are reported.