Guilherme Palermo Coelho

Designing ensembles of fuzzy classification systems: an immune-inspired approach

In this work we propose an immune-based approach for designing of fuzzy systems. From numerical data and with membership function previously defined, the immune algorithm evolves a population of fuzzy classification rules based on the clonal selection, hypermutation and immune network principles. Once AIS are able to find multiple good solutions of the problem, accurate and diverse fuzzy systems are built in a single run. Hence, we construct an ensemble of these classifier in order to achieve better results. An ensemble of classifiers consists of a set of individual classifiers whose outputs are combined when classifying novel patterns. The good performance of an ensemble is strongly dependent of individual accuracy and diversity of its components. We evaluate the proposed methodology through computational experiments on some datasets. The results demonstrate that the performance of the obtained fuzzy systems in isolation is very good. However when we combine these systems, a significant improvement is obtained in the correct classification rate, outperforming the single best classifier.

Omni-aiNet: an immune-inspired approach for omni optimization

This work presents omni-aiNet, an immune-inspired algorithm developed to solve single and multi-objective optimization problems, either with single and multi-global solutions. The search engine is capable of automatically adapting the exploration of the search space according to the intrinsic demand of the optimization problem. This proposal unites the concepts of omni-optimization, already proposed in the literature, with distinctive procedures associated with immune-inspired concepts. Due to the immune inspiration, the omni-aiNet presents a population capable of adjusting its size during the execution of the algorithm, according to a predefined suppression threshold, and a new grid mechanism to control the spread of solutions in the objective space. The omni-aiNet was applied to several optimization problems and the obtained results are presented and analyzed.

A Concentration-based Artificial Immune Network for continuous optimization

Metaheuristics based on the Artificial Immune System (AIS) framework, especially those inspired by the Immune Network theory, are known to be capable of stimulating the generation of diverse sets of solutions for a given problem, even though they generally implement very simple mechanisms to control the dynamics of the network. In the AIS literature, several studies propose different models that try to explain the behavior of immune networks, which are generally based on the concentration of antibodies and tend to better mimic some aspects of such complex systems. Therefore, in this work we propose a novel immune-inspired algorithm for optimization, named cob-aiNet (Concentration-based Artificial Immune Network), that intends to explore such network models and introduce new mechanisms to better control the dynamics of the network, so that a broader coverage of promising regions of the search space can be achieved. This property of cob-aiNet was verified in experimental analyses, in which the algorithm was compared to two other AIS proposals and also to all the competitors from the 2005 CEC Special Session on RealParameter Optimization.

Conceptual and Practical Aspects of the aiNet Family of Algorithms

In this paper, a review of the conceptual and practical aspects of the aiNet (Artificial Immune Network) family of algorithms will be provided. This family of algorithms started with the aiNet algorithm, proposed in 2002 for data clustering and, since then, several variations have been developed for data clustering, biclustering and optimization in general. Although the algorithms will be positioned with respect to other pertinent approaches from the literature, the emphasis of this paper will be on the formalization and critical analysis of the set of contributions produced along almost one decade of research in this specific theme, together with the provision of insights for further extensions.

Multi-Objective Biclustering: When Non-dominated Solutions are not Enough

The biclustering technique was developed to avoid some of the drawbacks presented by standard clustering techniques, such as their impossibility of finding correlating data under a subset of features, and, consequently, to allow the extraction of more accurate information from datasets. Given that biclustering requires the optimization of at least two conflicting objectives (residue and volume) and that multiple independent solutions are desirable as the outcome, a few multi-objective evolutionary algorithms for biclustering were proposed in the literature. However, these algorithms only focus their search in the generation of a global set of non-dominated biclusters, which may be insufficient for most of the problems as the coverage of the dataset can be compromised. In order to overcome such problem, a multi-objective artificial immune system capable of performing a multipopulation search, named MOM-aiNet, was proposed. In this work, the MOM-aiNet algorithm will be described in detail, and an extensive set of experimental comparisons will be performed, with the obtained results of MOM-aiNet being confronted with those produced by the popular CC algorithm, by another immune-inspired approach for biclustering (BIC-aiNet), and by the multi-objective approach for biclustering proposed by Mitra & Banka.

A Multi-Objective Multipopulation Approach for Biclustering

Biclustering is a technique developed to allow simultaneous clustering of rows and columns of a dataset. This might be useful to extract more accurate information from sparse datasets and to avoid some of the drawbacks presented by standard clustering techniques, such as their impossibility of finding correlating data under a subset of features. Given that biclustering requires the optimization of two conflicting objectives (residue and volume) and that multiple independent solutions are desirable as the outcome, a multi-objective artificial immune system capable of performing a multipopulation search, named MOM-aiNet, will be proposed in this paper. To illustrate the capabilities of this novel algorithm, MOM-aiNet was applied to the extraction of biclusters from two datasets, one taken from a well-known gene expression problem and the other from a collaborative filtering application. A comparative analysis has also been accomplished, with the obtained results being confronted with the ones produced by two popular biclustering algorithms from the literature (FLOC and CC) and also by another immune-inspired approach for biclustering (BIC-aiNet).

The Influence of the Pool of Candidates on the Performance of Selection and Combination Techniques in Ensembles

In this paper, we propose the use of an immune-inspired approach called opt-aiNet to generate a diverse set of high-performance candidates to compose an ensemble of neural network classifiers. Being a population-based search algorithm, the opt-aiNet is capable of maintaining diversity and finding many high-performance solutions simultaneously, which are known to be desired features when synthesizing an ensemble. Concerning the selection and combination phases, the most relevant selection and combination techniques already proposed in the literature have been considered. The main contribution of this paper is the indication that there is no pair of selection/combination technique that can be considered the best one, because the performance of the obtained ensemble varies significantly with the current composition of the pool of candidates already produced by the generation phase. Notwithstanding, this variability in performance is not restricted to the choice of opt-aiNet as the generative device. As a consequence, to overcome the performance of the best individual classifier, every possible pairs of selection and combination techniques should be tried. Only with such an exhaustive search (notice that the main computational burden is usually related to the generation phase), the performance of the ensemble was invariably superior to the performance of the best individual classifier on four benchmark classification problems.

Multivariate ant colony optimization in continuous search spaces

This work introduces an ant-inspired algorithm for optimization in continuous search spaces that is based on the generation of random vectors with multivariate Gaussian pdf. The proposed approach is called MACACO -- Multivariate Ant Colony Algorithm for Continuous Optimization -- and is able to simultaneously adapt all the dimensions of the random distribution employed to generate the new individuals at each iteration. In order to analyze MACACOs search efficiency, the approach was compared to a pair of counterparts: the Continuous Ant Colony System (CACS) and the approach known as Ant Colony Optimization in en (ACOR). The comparative analysis, which involves well-known benchmark problems from the literature, has indicated that MACACO outperforms CACS and ACOR in most cases as the quality of the final solution is concerned, and it is just about two times more costly than the least expensive contender.

Selection of Representative Models for Decision Analysis Under Uncertainty

The decision-making process in oil fields includes a step of risk analysis associated with the uncertainties present in the variables of the problem. Such uncertainties lead to hundreds, even thousands, of possible scenarios that are supposed to be analyzed so an effective production strategy can be selected. Given this high number of scenarios, a technique to reduce this set to a smaller, feasible subset of representative scenarios is imperative. The selected scenarios must be representative of the original set and also free of optimistic and pessimistic bias. This paper is devoted to propose an assisted methodology to identify representative models in oil fields. To do so, first a mathematical function was developed to model the representativeness of a subset of models with respect to the full set that characterizes the problem. Then, an optimization tool was implemented to identify the representative models of any problem, considering not only the cross-plots of the main output variables, but also the risk curves and the probability distribution of the attribute-levels of the problem. The proposed technique was applied to two benchmark cases and the results, evaluated by experts in the field, indicate that the obtained solutions are richer than those identified by previously adopted manual approaches. The program bytecode is available under request. HighlightsA new optimization-based method to select representative models in oil fields.A new mathematical function that captures the representativeness of a set of models.The mathematical function is combined with an optimization metaheuristic.The proposal was applied to the UNISIM-I-D benchmark problem to validate the methodology.Experts indicate that results are richer than those obtained by other approaches.

bicACO: An Ant Colony Inspired Biclustering Algorithm

A recent proposal developed to avoid some of the drawbacks presented by standard clustering algorithms is the so-called biclusteringtechnique [1], which performs clustering of rows and columns of the data matrix simultaneously, allowing the extraction of additional information from the dataset. Since the biclustering problem is combinatorial, and ant-based systems present several advantages when dealing with this kind of problems [2], in this work we propose an antinspired algorithm for biclustering, which was named bicACO.