Learning cluster-based classification systems with ant colony optimization algorithms

Classification is a data mining task the goal of which is to learn a model, from a training dataset, that can predict the class of a new data instance, while clustering aims to discover natural instance-groupings within a given dataset. Learning cluster-based classification systems involves partitioning a training set into data subsets (clusters) and building a local classification model for each data cluster. The class of a new instance is predicted by first assigning the instance to its nearest cluster and then using that cluster’s local classification model to predict the instance’s class. In this paper, we present an ant colony optimization (ACO) approach to building cluster-based classification systems. Our ACO approach optimizes the number of clusters, the positioning of the clusters, and the choice of classification algorithm to use as the local classifier for each cluster. We also present an ensemble approach that allows the system to decide on the class of a given instance by considering the predictions of all local classifiers, employing a weighted voting mechanism based on the fuzzy degree of membership in each cluster. Our experimental evaluation employs five widely used classification algorithms: naïve Bayes, nearest neighbour, Ripper, C4.5, and support vector machines, and results are reported on a suite of 54 popular UCI benchmark datasets.     

An experimental analysis of design choices of multi-objective ant colony optimization algorithms

There have been several proposals on how to apply the ant colony optimization (ACO) metaheuristic to multi-objective combinatorial optimization problems (MOCOPs). This paper proposes a new formulation of these multi-objective ant colony optimization (MOACO) algorithms. This formulation is based on adding specific algorithm components for tackling multiple objectives to the basic ACO metaheuristic. Examples of these components are how to represent multiple objectives using pheromone and heuristic information, how to select the best solutions for updating the pheromone information, and how to define and use weights to aggregate the different objectives. This formulation reveals more similarities than previously thought in the design choices made in existing MOACO algorithms. The main contribution of this paper is an experimental analysis of how particular design choices affect the quality and the shape of the Pareto front approximations generated by each MOACO algorithm. This study provides general guidelines to understand how MOACO algorithms work, and how to improve their design.     

Medoid-based clustering using ant colony optimization

The application of ACO-based algorithms in data mining has been growing over the last few years, and several supervised and unsupervised learning algorithms have been developed using this bio-inspired approach. Most recent works about unsupervised learning have focused on clustering, showing the potential of ACO-based techniques. However, there are still clustering areas that are almost unexplored using these techniques, such as medoid-based clustering. Medoid-based clustering methods are helpful—compared to classical centroid-based techniques—when centroids cannot be easily defined. This paper proposes two medoid-based ACO clustering algorithms, where the only information needed is the distance between data: one algorithm that uses an ACO procedure to determine an optimal medoid set (METACOC algorithm) and another algorithm that uses an automatic selection of the number of clusters (METACOC-K algorithm). The proposed algorithms are compared against classical clustering approaches using synthetic and real-world datasets.     

Learning Bayesian network classifiers using ant colony optimization

Bayesian networks are knowledge representation tools that model the (in)dependency relationships among variables for probabilistic reasoning. Classification with Bayesian networks aims to compute the class with the highest probability given a case. This special kind is referred to as Bayesian network classifiers. Since learning the Bayesian network structure from a dataset can be viewed as an optimization problem, heuristic search algorithms may be applied to build high-quality networks in medium- or large-scale problems, as exhaustive search is often feasible only for small problems. In this paper, we present our new algorithm, ABC-Miner, and propose several extensions to it. ABC-Miner uses ant colony optimization for learning the structure of Bayesian network classifiers. We report extended computational results comparing the performance of our algorithm with eight other classification algorithms, namely six variations of well-known Bayesian network classifiers, cAnt-Miner for discovering classification rules and a support vector machine algorithm.     

Reconstruction of phylogenetic trees using the ant colony optimization paradigm

We developed a new approach for the reconstruction of phylogenetic trees using ant colony optimization metaheuristics. A tree is constructed using a fully connected graph and the problem is approached similarly to the well-known traveling salesman problem. This methodology was used to develop an algorithm for constructing a phylogenetic tree using a pheromone matrix. Two data sets were tested with the algorithm: complete mitochondrial genomes from mammals and DNA sequences of the p53 gene from several eutherians. This new methodology was found to be superior to other well-known softwares, at least for this data set. These results are very promising and suggest more efforts for further developments.     

A critical analysis of parameter adaptation in ant colony optimization

Applying parameter adaptation means operating on parameters of an algorithm while it is tackling an instance. For ant colony optimization, several parameter adaptation methods have been proposed. In the literature, these methods have been shown to improve the quality of the results achieved in some particular contexts. In particular, they proved to be successful when applied to novel ant colony optimization algorithms for tackling problems that are not a classical testbed for optimization algorithms. In this paper, we show that the adaptation methods proposed so far do not improve, and often even worsen the performance when applied to high performing ant colony optimization algorithms for some classical combinatorial optimization problems.     

Incorporating heuristic information into ant colony optimization for epistasis detection

Epistasis has been receiving increasing attention in understanding the mechanism underlying susceptibility to complex diseases. Though many works have been done for epistasis detection, genome-wide association study remains a challenging task: it makes the search space excessively huge while solution quality is excessively demanded. In this study, we introduce an ant colony optimization based algorithm, AntMiner, by incorporating heuristic information into ant-decision rules. The heuristic information is used to direct ants in the search process for improving computational efficiency and solution accuracy. During iterations, chi-squared test is conducted to measure the association between an interaction and the phenotype. At the completion of the iteration process, statistically significant epistatic interactions are ordered and then screened by a post-procedure. Experiments of AntMiner and its comparison with existing algorithms epiMODE, TEAM and AntEpiSeeker are performed on both simulation data sets and real age-related macular degeneration data set, under the criteria of detection power and sensitivity. Results demonstrate that AntMiner is promising for epistasis detection. In terms of detection power, AntMiner performs best among all the other algorithms on all cases regardless of epistasis models and single nucleotide polymorphism size; compared with AntEpiSeeker, AntMiner can obtain better detection power but with less ants and iterations. In terms of sensitivity, AntMiner is better than AntEpiSeeker in detecting epistasis models displaying marginal effects but it has moderate sensitivity on epistasis models displaying no marginal effects. The study may provide clues on heuristics for further epistasis detection. The software package is available online at https://sourceforge.net/projects/antminer/files/.     

Dynamic load balancing on heterogeneous clusters for parallel ant colony optimization

Ant colony optimisation (ACO) is a nature-inspired, population-based metaheuristic that has been used to solve a wide variety of computationally hard problems. In order to take full advantage of the inherently stochastic and distributed nature of the method, we describe a parallelization strategy that leverages these features on heterogeneous and large-scale, massively-parallel hardware systems. Our approach balances workload effectively, by dynamically assigning jobs to heterogeneous resources which then run ACO implementations using different search strategies. Our experimental results confirm that we can obtain significant improvements in terms of both solution quality and energy expenditure, thus opening up new possibilities for the development of metaheuristic-based solutions to “real world” problems on high-performance, energy-efficient contemporary heterogeneous computing platforms.     

The Markovian binary tree applied to demography

We apply matrix analytic methods and branching processes theory to a comparison of female populations in different countries. We show how the same mathematical model allows us to determine characteristics about individual women, such as the distribution of her lifetime, the time until her first and her last daughter, and the number of daughters, as well as to analyze properties of the whole female family generated by a first woman, such as the extinction probability of the family, the distributions of the time until extinction, of the family size at any given time and of the total progeny.     

Interactive ant colony optimization (iACO) for early lifecycle software design

Finding good designs in the early stages of the software development lifecycle is a demanding multi-objective problem that is crucial to success. Previously, both interactive and non-interactive techniques based on evolutionary algorithms (EAs) have been successfully applied to assist the designer. However, recently ant colony optimization was shown to outperform EAs at optimising quantitative measures of software designs with a limited computational budget. In this paper, we propose a novel interactive ACO (iACO) approach, in which the search is steered jointly by an adaptive model that combines subjective and objective measures. Results show that iACO is speedy, responsive and effective in enabling interactive, dynamic multi-objective search. Indeed, study participants rate the iACO search experience as compelling. Moreover, inspection of the learned model facilitates understanding of factors affecting users’ judgements, such as the interplay between a design’s elegance and the interdependencies between its components.     

Estimation-based ant colony optimization and local search for the probabilistic traveling salesman problem

The use of ant colony optimization for solving stochastic optimization problems has received a significant amount of attention in recent years. In this paper, we present a study of enhanced ant colony optimization algorithms for tackling a stochastic optimization problem, the probabilistic traveling salesman problem. In particular, we propose an empirical estimation approach to evaluate the cost of the solutions constructed by the ants. Moreover, we use a recent estimation-based iterative improvement algorithm as a local search. Experimental results on a large number of problem instances show that the proposed ant colony optimization algorithms outperform the current best algorithm tailored to solve the given problem, which also happened to be an ant colony optimization algorithm. As a consequence, we have obtained a new state-of-the-art ant colony optimization algorithm for the probabilistic traveling salesman problem.     

An ant colony optimization algorithm for phylogenetic estimation under the minimum evolution principle

Background  

Distance matrix methods constitute a major family of phylogenetic estimation methods, and the minimum evolution (ME) principle (aiming at recovering the phylogeny with shortest length) is one of the most commonly used optimality criteria for estimating phylogenetic trees. The major difficulty for its application is that the number of possible phylogenies grows exponentially with the number of taxa analyzed and the minimum evolution principle is known to belong to the -hard class of problems.     

Spatial dynamics of colony interactions in young populations of the fire ant Solenopsis invicta

Newly founded colonies of the fire ant Solenopsis invicta compete intensely by brood raids, which result in a rapid reduction of colony density. Experimental plantings of colonies and analyses of sequential maps were used to examine the importance of spatial pattern in the dynamics of young populations. Colony positions were initially clumped in naturally founded cohorts, but were regular in most mature populations. Incipient colonies planted in clumped patterns were more likely to engage in brood raids than colonies planted in regular hexagonal patterns at the same average density. However, contrary to what would be expected if local crowding increased mortality, no significant increases in spatial regularity were observed during brood raiding either in the experimentally planted populations or in a natural population of more than 1200 incipient colonies. These results show that it may be difficult to infer the degree of past or current competition by passive analysis of spatial data even when field experiments show that the probability of mortality depends on local spacing.     

The sampling properties of some distance geometry algorithms applied to unconstrained polypeptide chains: a study of 1830 independently computed conformations

In this paper we study the statistical geometry of ensembles of poly (L-alanine) conformations computed by several different distance geometry algorithms. Since basic theory only permits us to predict the statistical properties of such ensembles a priori when the distance constraints have a very simple form, the only constraints used for these calculations are those necessary to obtain reasonable bond lengths and angles, together with a lack of short- and long-range atomic overlaps. The geometric properties studied include the squared end-to-end distance and radius of gyration of the computed conformations, in addition to the usual rms coordinate and phi/psi angle deviations among these conformations. The distance geometry algorithms evaluated include several variations of the well-known embed algorithm, together with optimizations of the torsion angles using the ellipsoid and variable target function algorithms. The conclusions may be summarized as follows: First, the distribution with which the trial distances are chosen in most implementations of the embed algorithm is not appropriate when no long-range upper bounds on the distances are present, because it leads to unjustifiably expanded conformations. Second, chosing the trial distances independently of one another leads to a lack of variation in the degree of expansion, which in turn produces a relatively low rms square coordinate difference among the members of the ensemble. Third, when short-range steric constraints are present, torsion angle optimizations that start from conformations obtained by choosing their phi/psi angles randomly with a uniform distribution between -180 degrees and +180 degrees do not converge to conformations whose angles are uniformly distributed over the sterically allowed regions of the phi/psi plane. Finally, in an appendix we show how the sampling obtained with the embed algorithm can be substantially improved upon by the proper application of existing methodology.     

ACOHAP: an efficient ant colony optimization for the haplotype inference by pure parsimony problem

Haplotype information plays an important role in many genetic analyses. However, the identification of haplotypes based on sequencing methods is both expensive and time consuming. Current sequencing methods are only efficient to determine conflated data of haplotypes, that is, genotypes. This raises the need to develop computational methods to infer haplotypes from genotypes.Haplotype inference by pure parsimony is an NP-hard problem and still remains a challenging task in bioinformatics. In this paper, we propose an efficient ant colony optimization (ACO) heuristic method, named ACOHAP, to solve the problem. The main idea is based on the construction of a binary tree structure through which ants can travel and resolve conflated data of all haplotypes from site to site. Experiments with both small and large data sets show that ACOHAP outperforms other state-of-the-art heuristic methods. ACOHAP is as good as the currently best exact method, RPoly, on small data sets. However, it is much better than RPoly on large data sets. These results demonstrate the efficiency of the ACOHAP algorithm to solve the haplotype inference by pure parsimony problem for both small and large data sets.     

The Krylov-Bogoliubov-Mitropolskii method applied to models of population dynamics

In this study the possibility of applying the asymptotic method of Krylov-Bogoliubov-Mitropolskii to problems of population dynamics is shown. Especially a general Volterra-Gause-Witt type model for prey-predator interaction is investigated. A discussion on the results obtained is given for the general model and for a particular case as well.