G. Pamparà | University of Pretoria (original) (raw)

Papers by G. Pamparà

Recently, continuous-valued optimization problems have received a great amount of focus, resultin... more Recently, continuous-valued optimization problems have received a great amount of focus, resulting in optimization algorithms which are very efficient within the continuousvalued space. Many optimization problems are, however, defined within the binaryvalued problem space. These continuous-valued optimization algorithms can not operate directly on a binary-valued problem representation, without algorithm adaptations because the mathematics used within these algorithms generally fails within a binary problem space. Unfortunately, such adaptations may alter the behavior of the algorithm, potentially degrading the performance of the original continuous-valued optimization algorithm. Additionally, binary representations present complications with respect to increasing problem dimensionality, interdependencies between dimensions, and a loss of precision. This research investigates the possiblity of applying continuous-valued optimization algorithms to solve binary-valued problems, without requiring algorithm adaptation. This is achieved through the application of a mapping technique, known as angle modulation. Angle modulation effectively addresses most of the problems associated with the use of a binary representation by abstracting a binary problem into a four-dimensional continuous-valued space, from which a binary solution is then obtained. The abstraction is obtained as a bit-generating function produced by a continuous-vaued algorithm. A binary solution is then obtained by sampling the bit-generating function. This thesis proposes a number of population-based angle-modulated continuousvalued algorithms to solve binary-valued problems. These algorithms are then compared to binary algorithm counterparts, using a suite of benchmark functions. Empirical analysis will show that the angle-modulated continuous-valued algorithms are viable alternatives to binary optimization algorithms.

Swarm and Evolutionary Computation, 2021

Abstract Quantification of the performance of algorithms that solve dynamic optimization problems... more Abstract Quantification of the performance of algorithms that solve dynamic optimization problems (DOPs) is challenging, since the fitness landscape changes over time. Popular performance measures for DOPs do not adequately account for ongoing fitness landscape scale changes, and often yield a confounded view of performance. Similarly, most popular measures do not allow for fair performance comparisons across multiple instances of the same problem type nor across different types of problems, since performance values are not normalized. Many measures also assume normally distributed input data values, while in reality the necessary conditions for data normality are often not satisfied. The majority of measures also fail to capture the notion of performance variance over time. This paper proposes a new performance measure for DOPs, namely the relative error distance. The measure shows how close to optimal an algorithm performs by considering the multi-dimensional distance between the vector comprising the normalized performance scores for specific algorithm iterations of interest, and the theoretical point of best possible performance. The new measure does not assume normally distributed performance data across fitness landscape changes, is resilient against fitness landscape scale changes, better incorporates performance variance across fitness landscape changes into a single scalar value, and allows easier algorithm comparisons using established nonparametric statistical methods.

Proceedings of the Genetic and Evolutionary Computation Conference Companion, 2019

Reproducible experimental work is a vital part of the scientific method. It is a concern that is ... more Reproducible experimental work is a vital part of the scientific method. It is a concern that is often, however, overlooked in modern computational intelligence research. Scientific research within the areas of programming language theory and mathematics have made advances that are directly applicable to the research areas of evolutionary and swarm intelligence. Through the use of functional programming and the established abstractions that functional programming provides, it is possible to define the elements of evolutionary and swarm intelligence algorithms as compositional computations. These compositional blocks then compose together to allow the declarations of an algorithm, whilst considering the declaration as a "sub-program". These sub-programs may then be executed at a later time and provide the blueprints of the computation. Storing experimental results within a robust data-set file format, which is widely supported by analysis tools, provides additional flexibility and allows different analysis tools to access datasets in the same efficient manner. This paper presents an open-source software library for evolutionary and swarm-intelligence algorithms which allows the type-safe, compositional, monadic and functional declaration of algorithms while tracking and managing effects (e.g. usage of a random number generator) that directly influences the execution of an algorithm.

Proceedings of the Genetic and Evolutionary Computation Conference Companion, 2019

Dynamic constrained optimization problems (DCOPs) provide larger complexity for an optimization a... more Dynamic constrained optimization problems (DCOPs) provide larger complexity for an optimization algorithm by changing the problem landscape throughout the optimization process. Introducing constraints to an already changing dynamic environment increases the observed complexity of the problem space. Allowing such constraints to have irregular shapes which change along with the problem space itself provides an even greater level of complexity for an optimization algorithm. This paper proposes a function generator capable of creating dynamically constrained dynamic environments by extending the moving peaks benchmark (MPB) function generator. An analysis of the resulting environments produced by the generator is performed using fitness landscape analysis (FLA). A visual inspection of the resulting generated environments is also included.

Lecture Notes in Computer Science, 2018

The quantum-inspired particle swarm optimization (QPSO) algorithm has been developed to find and ... more The quantum-inspired particle swarm optimization (QPSO) algorithm has been developed to find and track an optimum for dynamic optimization problems. Though QPSO has been shown to be effective, despite its simplicity, it does introduce an additional control parameter: the radius of the quantum cloud. The performance of QPSO is sensitive to the value assigned to this problem dependent parameter, which basically limits the area of the search space wherein new, better optima can be detected. This paper proposes a strategy to dynamically adapt the quantum radius, with changes in the environment. A comparison of the adaptive radius QPSO with the static radius QPSO showed that the adaptive approach achieves desirable results, without prior tuning of the quantum radius.

2015 IEEE Symposium Series on Computational Intelligence, 2015

This paper proposes a library for computational intelligence using functional programming to addr... more This paper proposes a library for computational intelligence using functional programming to address the complexities in algorithm implementation and highlighting specific concerns that are often ignored in the algorithm descriptions. Useful abstractions, common in the paradigm of functional programming, are used to make implementation specifics of algorithms part of the algorithm definition, resulting in the tracking of these effects, together with the control of the effects. Effects, requiring management within an algorithm, include the use of pseudo-random number generators, writing data to files or the console, or providing the control parameter configuration of the algorithm in an experiment. By defining the units of work for an algorithm in a general and generic form, composition of these different algorithmic units is possible, thereby creating larger, more complex computational units. The software library providing such reusable, peer-reviewed compos able computational unit,...

IEEE SSCI 2011 - Symposium Series on Computational Intelligence - SIS 2011: 2011 IEEE Symposium on Swarm Intelligence, 2011

Artificial bee colony (ABC) optimization is a rela- tively new population-based, stochastic optim... more Artificial bee colony (ABC) optimization is a rela- tively new population-based, stochastic optimization technique. ABC was developed to optimize unconstrained problems within continuous-valued domains. This paper proposes three versions of ABC that enable it to be applied to optimization problems with binary-valued domains. The performances of these binary ABC algorithms are compared on a benchmark of unconstrained optimization problems. The best of these algorithms, i.e. angle- modulated ABC (AMABC), is then compared with the angle- modulated particle swarm optimizer and the angle-modulated differential evolution algorithm. I. INTRODUCTION Over the past decade, the growing interest in computational swarm intelligence has resulted in a number of new optimiza- tion algorithms. Of these, the ant colony meta-heuristic (1) and partile swarm optimization (PSO) (2) are well-known and studied. Recently, algorithms based on models of bee foraging behavior (3), fish schools (4), bacteria (5), and fireflies (6) were developed. This paper focuses on one of these recent swarm- based models, namely artificial bee colony (ABC) optimization as introduced by Karaboga (3). ABC is an algorithmic model of the foraging behavior of honey bees, and was developed to solve unconstrained optimization problems with continuous- valued domains. ABC has been shown to be an efficient optimizer for such problems (7).

2006 IEEE International Conference on Evolutionary Computation, 2000

The ability of differential evolution (DE) to perform well in continuous-valued search spaces is ... more The ability of differential evolution (DE) to perform well in continuous-valued search spaces is well documented. The arithmetic reproduction operator used by differential evolution is simple, however, the manner in which the operator is defined, makes it practically impossible to effectively apply the standard DE to other problem spaces. An interesting and unique mapping method is examined which will enable the DE algorithm to operate within binary space. Using angle modulation, a bit string can be generated using a trigonometric generating function. The DE is used to evolve the coefficients to the trigonometric function, thereby allowing a mapping from continuous-space to binary-space. Instead of evolving the higher-dimensional binary solution directly, angle modulation is used together with DE to reduce the complexity of the problem into a 4-dimensional continuous-valued problem. Experimental results indicate the effectiveness of the technique and the viability for the DE to operate in binary space.

Proceedings - 2005 IEEE Swarm Intelligence Symposium, SIS 2005, 2005

Niching algorithms have the ability to locate and maintain more than one solution to a multi-moda... more Niching algorithms have the ability to locate and maintain more than one solution to a multi-modal optimization problem. Recently, niching algorithms have been developed for particle swarm optimization (PSO) to locate multiple optima. This paper investigates the ability of the basic PSO to locate and maintain niches, in order to arrive at a conclusion on whether special purpose PSO algorithms,

Proceedings - 2005 IEEE Swarm Intelligence Symposium, SIS 2005, 2005

This paper presents CiClops, which is a virtual laboratory for performing experiments, using comp... more This paper presents CiClops, which is a virtual laboratory for performing experiments, using computational intelligence (CI) algorithms that scale over multiple workstations. Additionally, the paper introduces CIlib, which is an open source library of CI algorithms, currently containing mostly particle swarm optimization (PSO) and ant colony optimization (ACO) algorithms. The main purpose of CiClops is to specify CI algorithms to

Proceedings of the International Joint Conference on Neural Networks, 2008

Research in computational intelligence (CI) has produced a huge collection of algorithms, grouped... more Research in computational intelligence (CI) has produced a huge collection of algorithms, grouped into the main CI paradigms. Development of a new CI algorithm requires such algorithm to be thoroughly benchmarked against existing algorithms, which requires researchers to implement already published algorithms. This re-implementation of existing algorithms unnecessarily wastes valuable time, and may be the cause of incorrect results due

2007 IEEE Congress on Evolutionary Computation, 2007

Differential evolution has shown to be a very powerful, yet simple, population-based optimization... more Differential evolution has shown to be a very powerful, yet simple, population-based optimization approach. The nature of its reproduction operator limits its application to continuous-valued search spaces. However, a simple discretization procedure can be used to convert floating-point solution vectors into discrete-valued vectors. This paper considers three approaches in which differential evolution can be used to solve problems with binary-valued parameters. The first approach is based on a homomorphous mapping [1], while the second approach interprets the floating-point solution vector as a vector of probabilities, used to decide on the appropriate binary value. The third approach normalizes solution vectors and then discretize these normalized vectors to form a bitstring. Empirical results are provided to illustrate the efficiency of both methods in comparison with particle swarm optimizers.

The optimisation process of a particular problem generally has many influencing factors including... more The optimisation process of a particular problem generally has many influencing factors including the parameter choices, problem constraints as well as the complexity of the optimistion algorithm and optimisation problem among others. The dimensionality of a problem influences the computational complexity in converging to a valid solution. With problems defined in larger and more abstract dimensions, complexity becomes a problem as the solutions presented by the algorithm are more likely to be sub-optimal. An interesting and unique manner to reduce the complexity of binary problems is developed in this paper: Angle modulation is applied to generate a bit string to solve binary problems, using Particle Swarm Optimisation (PSO) to evolve the function coefficients of a trigonometric model. Instead of evolving a high dimensional bit vector, angle modulation reduces the problem to a four-dimensional problem defined in continuous space. Experimental results show that the angle modulation method is faster than the standard Binary PSO, and that accuracy is improved for most benchmark functions used.

Proceedings - 2005 IEEE Swarm Intelligence Symposium, SIS 2005, 2005

IEEE SSCI 2011 - Symposium Series on Computational Intelligence - SIS 2011: 2011 IEEE Symposium on Swarm Intelligence, 2011

Swarm and Evolutionary Computation, 2021

Proceedings of the Genetic and Evolutionary Computation Conference Companion, 2019

Lecture Notes in Computer Science, 2018

2015 IEEE Symposium Series on Computational Intelligence, 2015

IEEE SSCI 2011 - Symposium Series on Computational Intelligence - SIS 2011: 2011 IEEE Symposium on Swarm Intelligence, 2011

2006 IEEE International Conference on Evolutionary Computation, 2000

Proceedings - 2005 IEEE Swarm Intelligence Symposium, SIS 2005, 2005

Proceedings of the International Joint Conference on Neural Networks, 2008

2007 IEEE Congress on Evolutionary Computation, 2007

Proceedings - 2005 IEEE Swarm Intelligence Symposium, SIS 2005, 2005

IEEE SSCI 2011 - Symposium Series on Computational Intelligence - SIS 2011: 2011 IEEE Symposium on Swarm Intelligence, 2011