Visual exploration of algorithm parameter space

Abstract

In this article we apply information visualization techniques to the domain of swarm intelligence. We describe an intuitive approach that enables researchers and designers of stochastic optimization algorithms to efficiently determine trends and identify optimal regions in an algorithm's parameter search space. The parameter space is evenly sampled using low-discrepancy sequences, and visualized using parallel coordinates. Various techniques are applied to iteratively highlight areas that influence the optimization algorithm's performance on a particular problem. By analyzing experimental data with this technique, we were able to gain important insight into the complexity of the target problem domain. For example, we were able to confirm some underlying theoretical assumptions of an important class of population-based stochastic algorithms. Most importantly, the technique improves the efficiency of finding good parameter settings by orders of magnitude.

Training multi-agent teams from zero knowledge with the competitive coevolutionary team-based particle swarm optimiser

“…Parameter optimisation was done using a parallel coordinate visualisation technique (Franken 2009). The parallel coordinate visualisation technique works by plotting the performance of different parameter combinations along with the measured performance of each parameter combination.…”

“…First developed by Kennedy and Eberhart (Kennedy and Eberhart 1995) in 1995, the PSO algorithm has been more successful in solving complex problems than traditional evolutionary computation (EC) algorithms (Kennedy and Eberhart 2001). Particle swarm optimisers have proved successful in training board state evaluators for games such as Tic-Tac-Toe, Checkers and Bao (Messerschmidt and Engelbrecht 2002;Franken andEngelbrecht 2003a,b, 2004;Conradie and Engelbrecht 2006). The aforementioned training techniques require the construction of traditional game trees, and using a competitive coevolutionary PSO to train a neural network game state evaluator.…”

Distributed Estimation of Distribution Algorithms for continuous optimization: How does the exchanged information influence their behavior?

“…With the purpose of offering an alternative view of the relationships between the parameters and the algorithm performance, the best and worst coordinates is a useful technique which has been successfully used to represent high dimensional data as polylines in two dimensions. More recently, it has been used to capture the underlying interactions between the parameters of a Particle Swarm Optimization algorithm [15]. In this paper, the first eight axes represent the parameters of the configurations, whereas the ninth one represents the Friedman average ranking mentioned before.…”

Set-based particle swarm optimization applied to the multidimensional knapsack problem

“…Only static control parameters were considered. In order to generate the parameter combinations in a manner that ensures that the parameter space was covered well, sequences of Sobol pseudo-random numbers were used according to the method proposed by Franken (2009). Even though the number of dimensions of the parameter space differs depending on the PSO algorithm, the same number of parameter combinations was used in tuning each of the algorithm-topology pairs on each of the problem sets.…”