Jorge Maturana

Autonomous operator management for evolutionary algorithms

The performance of an evolutionary algorithm strongly depends on the design of its operators and on the management of these operators along the search; that is, on the ability of the algorithm to balance exploration and exploitation of the search space. Recent approaches automate the tuning and control of the parameters that govern this balance. We propose a new technique to dynamically control the behavior of operators in an EA and to manage a large set of potential operators. The best operators are rewarded by applying them more often. Tests of this technique on instances of 3-SAT return results that are competitive with an algorithm tailored to the problem.

Extreme compass and Dynamic Multi-Armed Bandits for Adaptive Operator Selection

The goal of Adaptive Operator Selection is the on-line control of the choice of variation operators within Evolutionary Algorithms. The control process is based on two main components, the credit assignment, that defines the reward that will be used to evaluate the quality of an operator after it has been applied, and the operator selection mechanism, that selects one operator based on some operators qualities. Two previously developed Adaptive Operator Selection methods are combined here: Compass evaluates the performance of operators by considering not only the fitness improvements from parent to offspring, but also the way they modify the diversity of the population, and their execution time; Dynamic Multi-Armed Bandit proposes a selection strategy based on the well-known UCB algorithm, achieving a compromise between exploitation and exploration, while nevertheless quickly adapting to changes. Tests with the proposed method, called ExCoDyMAB, are carried out using several hard instances of the Satisfiability problem (SAT). Results show the good synergetic effect of combining both approaches.

Solving the short-term electrical generation scheduling problem by an adaptive evolutionary approach

In this paper, we introduce an adaptive evolutionary approach to solve the short-term electrical generation scheduling problem (STEGS). The STEGS is a hard constraint satisfaction optimization problem. The algorithm includes various strategies proposed in the literature to tackle hard problems with constraints such as: the representation used a non-binary coding scheme that drastically reduces the search space compared with the traditional evolutionary approaches. Specialized operators are especially designed for this problem and for this kind of representation, which also includes a local search procedure. Furthermore, the algorithm is guided by an adaptive parameter control strategy. We used some very well known benchmarks for STEGS to evaluate our approach. The results are very encouraging and we have obtained new better values for all the systems tested. Our aim here is to show that evolutionary approaches can be considered as good techniques to be used to solve real-world highly constrained problems.

On the design of adaptive control strategies for evolutionary algorithms

This paper focuses on the design of control strategies forEvolutionary Algorithms. We propose a method to encapsulate multipleparameters, reducing control to only one criterion. This method allowsto define generic control strategies independently from both the algorithmsoperators and the problem to be solved. Three strategies areproposed and compared on a classical optimization problem, consideringtheir generality and performance.

An exploration-exploitation compromise-based adaptive operator selection for local search

This paper deals with the adaptive selection of operators in the context of local search (LS). In evolutionary algorithms, diversity is a key concept. We consider a related idea: the similarity between the candidate solution and the solutions in the search trajectory. This notion, together with the solution quality, is used to evaluate the performance of each operator. A new utility measure for LS operators, evaluating relative distances between the operators, is introduced. It is compared with an existing measure based on the Pareto dominance relationship using some basic selection schemes. An adaptive version of the algorithm is also examined. The proposed methods are tested on the Quadratic Assignment Problem and Asymmetric Traveling Salesman Problem.

Adaptive Operator Selection and Management in Evolutionary Algorithms

One of the settings that most affect the performance of Evolutionary Algorithms is the selection of the variation operators that are efficient to solve the problem at hand. The control of these operators can be handled in an autonomous way, while solving the problem, at two different levels: at the structural level, when deciding which operators should be part of the algorithmic framework, referred to as Adaptive Operator Management (AOM); and at the behavioral level, when selecting which of the available operators should be applied at a given time instant, called as Adaptive Operator Selection (AOS). Both controllers guide their choices based on a common knowledge about the recent performance of each operator. In this chapter, we present methods for these two complementary aspects of operator control, the ExCoDyMAB AOS and the Blacksmith AOM, providing case studies to analyze them in order to highlight the major issues that should be considered for the design of more autonomous Evolutionary Algorithms.

Controlling behavioral and structural parameters in evolutionary algorithms

Evolutionary algorithms have been efficiently used for solving combinatorial problems. However a successful application rely on a good definition of the algorithm structure and a correct search guidance. Similarly to the majority of metaheuristic methods, the performance of an evolutionary algorithm is intrinsically linked to its ability to properly manage the balance between the exploitation and the exploration of the search space. Recently, new approaches have emerged to make these algorithms more independent, especially by automating the setting of parameters. We propose a new approach whose objective is twofold: (1) to manage an important set of potential operators, whose performances are a priori unknown, and (2) to dynamically control the behavior of operators in a evolutionary algorithm, thanks to their probabilities of

A comparison of operator utility measures for on-line operator selection in local search

This paper investigates the adaptive selection of operators in the context of Local Search. The utility of each operator is computed from the solution quality and distance of the candidate solution from the search trajectory. A number of utility measures based on the Pareto dominance relationship and the relative distances between the operators are proposed and evaluated on QAP instances using an implied or static target balance between exploitation and exploration. A refined algorithm with an adaptive target balance is then examined.

From adaptive to more dynamic control in evolutionary algorithms

Adaptive evolutionary algorithms have been widely developed to improve the management of the balance between intensification and diversification during the search. Nevertheless, this balance may need to be dynamically adjusted over time. Based on previous works on adaptive operator selection, we investigate in this paper how an adaptive controller can be used to achieve more dynamic search scenarios and what is the real impact of possible combinations of control components. This study may be helpful for the development of more autonomous and efficient evolutionary algorithms.

Robust optimization by means of vegetative reproduction

In the last half century, computer science has witnessed the appearance of nature-inspired methods for the resolution of complex optimization problems, which are hardly solved by traditional optimization methods. Metaheuristics like evolutionary algorithms or swarm intelligence have been successfully applied to a wide range of both theoretical and practical problems. This article presents a new optimization method based in reproduction mechanics of plant clonal colonies. These systems are composed of a set of clones, interconnected and spatially spread over a geographical area. In this new metaheuristic, called Clonal Colony Optimization (CCO), problem solutions are associated to clones, that are subject to evolutionary cycles that adaptively reconfigure the geographical covering over the search space of the problem. Solutions coded in this manner would be more robust that those obtained using independent individuals.