Skip to main content

Research Repository

Advanced Search

All Outputs (32)

A Deep Dive into Effects of Structural Bias on CMA-ES Performance along Affine Trajectories (2024)
Presentation / Conference Contribution
van Stein, N., Thomson, S. L., & Kononova, A. V. (2024, September). A Deep Dive into Effects of Structural Bias on CMA-ES Performance along Affine Trajectories. Paper presented at Parallel Problem Solving from Nature (PPSN) 2024, Hagenberg, Austria

To guide the design of better iterative optimisation heuristics, it is imperative to understand how inherent structural biases within algorithm components affect the performance on a wide variety of search landscapes. This study explores the impact o... Read More about A Deep Dive into Effects of Structural Bias on CMA-ES Performance along Affine Trajectories.

Entropy, Search Trajectories, and Explainability for Frequency Fitness Assignment (2024)
Presentation / Conference Contribution
Thomson, S. L., Ochoa, G., van den Berg, D., Liang, T., & Weise, T. (2024, September). Entropy, Search Trajectories, and Explainability for Frequency Fitness Assignment. Presented at Parallel Problem Solving from Nature (PPSN 2024), Hagenberg, Austria

Local optima are a menace that can trap optimisation processes. Frequency fitness assignment (FFA) is an concept aiming to overcome this problem. It steers the search towards solutions with rare fitness instead of high-quality fitness. FFA-based algo... Read More about Entropy, Search Trajectories, and Explainability for Frequency Fitness Assignment.

Explaining evolutionary feature selection via local optima networks (2024)
Presentation / Conference Contribution
Adair, J., Thomson, S. L., & Brownlee, A. E. (2024, July). Explaining evolutionary feature selection via local optima networks. Presented at ACM Genetic and Evolutionary Computation Conference (GECCO) 2024, Melbourne, Australia

We analyse fitness landscapes of evolutionary feature selection to obtain information about feature importance in supervised machine learning. Local optima networks (LONs) are a compact representation of a landscape, and can potentially be adapted fo... Read More about Explaining evolutionary feature selection via local optima networks.

Exploring the use of fitness landscape analysis for understanding malware evolution (2024)
Presentation / Conference Contribution
Babaagba, K., Murali, R., & Thomson, S. L. (2024, July). Exploring the use of fitness landscape analysis for understanding malware evolution. Presented at ACM Genetic and Evolutionary Computation Conference (GECCO) 2024, Melbourne, Australia

We conduct a preliminary study exploring the potential of using fitness landscape analysis for understanding the evolution of malware. This type of optimisation is fairly new and has not previously been studied through the lens of landscape analysis.... Read More about Exploring the use of fitness landscape analysis for understanding malware evolution.

Where the Really Hard Quadratic Assignment Problems Are: the QAP-SAT instances (2024)
Presentation / Conference Contribution
Verel, S., Thomson, S. L., & Rifki, O. (2024, April). Where the Really Hard Quadratic Assignment Problems Are: the QAP-SAT instances. Presented at EvoCOP 2024, Aberystwyth, UK

The Quadratic Assignment Problem (QAP) is one of the major domains in the field of evolutionary computation, and more widely in combinatorial optimization. This paper studies the phase transition of the QAP, which can be described as a dramatic chang... Read More about Where the Really Hard Quadratic Assignment Problems Are: the QAP-SAT instances.

Frequency Fitness Assignment for Untangling Proteins in 2D (2024)
Presentation / Conference Contribution
Koutstaal, J., Kommandeur, J., Timmer, R., Horn, R., Thomson, S. L., & van den Berg, D. (2024, April). Frequency Fitness Assignment for Untangling Proteins in 2D. Presented at EvoStar 2024, Aberyswyth, UK

At the time of writing, there is no known deterministic-time algorithm to sample valid initial solutions with uniform random distribution for the HP protein folding model, because guaranteed uniform random sampling produces collisions (i.e. constrain... Read More about Frequency Fitness Assignment for Untangling Proteins in 2D.

Understanding fitness landscapes in morpho-evolution via local optima networks (2024)
Presentation / Conference Contribution
Thomson, S. L., Le Goff, L., Hart, E., & Buchanan, E. (2024, July). Understanding fitness landscapes in morpho-evolution via local optima networks. Presented at Genetic and Evolutionary Computation Conference (GECCO 2024), Melbourne, Australia

Morpho-Evolution (ME) refers to the simultaneous optimisation of a robot's design and controller to maximise performance given a task and environment. Many genetic encodings have been proposed which are capable of representing design and control. Pre... Read More about Understanding fitness landscapes in morpho-evolution via local optima networks.

Temporal True and Surrogate Fitness Landscape Analysis for Expensive Bi-Objective Optimisation Expensive Bi-Objective (2024)
Presentation / Conference Contribution
Rodriguez, C. J., Thomson, S. L., Alderliesten, T., & Bosman, P. A. N. (2024, July). Temporal True and Surrogate Fitness Landscape Analysis for Expensive Bi-Objective Optimisation Expensive Bi-Objective. Presented at Genetic and Evolutionary Computation Conference (GECCO 2024), Melbourne, Australia

Many real-world problems have expensive-to-compute fitness functions and are multi-objective in nature. Surrogate-assisted evolutionary algorithms are often used to tackle such problems. Despite this, literature about analysing the fitness landscapes... Read More about Temporal True and Surrogate Fitness Landscape Analysis for Expensive Bi-Objective Optimisation Expensive Bi-Objective.

Information flow and Laplacian dynamics on local optima networks (2024)
Presentation / Conference Contribution
Richter, H., & Thomson, S. L. (2024, June). Information flow and Laplacian dynamics on local optima networks. Presented at IEEE Congress on Evolutionary Computation (IEEE CEC), Yokohama, Japan

We propose a new way of looking at local optima networks (LONs). LONs represent fitness landscapes; the nodes are local optima, and the edges are search transitions between them. Many metrics computed on LONs have been proposed and shown to be linked... Read More about Information flow and Laplacian dynamics on local optima networks.

The Opaque Nature of Intelligence and the Pursuit of Explainable AI (2023)
Presentation / Conference Contribution
Thomson, S. L., van Stein, N., van den Berg, D., & van Leeuwen, C. (2023, November). The Opaque Nature of Intelligence and the Pursuit of Explainable AI. Presented at NCTA 2023: 15th International Conference on Neural Computation Theory and Applications, Rome, Italy

When artificial intelligence is used for making decisions, people are more likely to accept those decisions if they can be made intelligible to the public. This understanding has led to the emerging field of explainable artificial intelligence. We re... Read More about The Opaque Nature of Intelligence and the Pursuit of Explainable AI.

Can HP-protein folding be solved with genetic algorithms? Maybe not (2023)
Presentation / Conference Contribution
Jansen, R., Horn, R., van Eck, O., Version, K., Thomson, S. L., & van den Berg, D. (2023, November). Can HP-protein folding be solved with genetic algorithms? Maybe not. Presented at ECTA 2023 15th International Conference on Evolutionary Computation Theory and Applications, Rome, Italy

Genetic algorithms might not be able to solve the HP-protein folding problem because creating random individuals for an initial population is very hard, if not impossible. The reason for this, is that the expected number of constraint violations incr... Read More about Can HP-protein folding be solved with genetic algorithms? Maybe not.

Too Constrained for Genetic Algorithms. Too Hard for Evolutionary Computing. The Traveling Tournament Problem. (2023)
Presentation / Conference Contribution
Verduin, K., Thomson, S. L., & van den Berg, D. (2023, November). Too Constrained for Genetic Algorithms. Too Hard for Evolutionary Computing. The Traveling Tournament Problem. Presented at ECTA 2023 15th International Conference on Evolutionary Computation Theory and Applications, Rome, Italy

Unlike other NP-hard problems, the constraints on the traveling tournament problem are so pressing that it’s hardly possible to randomly generate a valid solution, for example, to use in a genetic algorithm’s initial population. In this study, we ran... Read More about Too Constrained for Genetic Algorithms. Too Hard for Evolutionary Computing. The Traveling Tournament Problem..

Unexplained Fluctuations in Particle Swarm Optimisation Performance with Increasing Problem Dimensionality (2023)
Presentation / Conference Contribution
Graham, K. C., Thomson, S. L., & Brownlee, A. E. I. (2023). Unexplained Fluctuations in Particle Swarm Optimisation Performance with Increasing Problem Dimensionality. . https://doi.org/10.1145/3583133.3596433

We study the behaviour of particle swarm optimisation (PSO) with increasing problem dimension for the Alpine 1 function as an exploratory and preliminary case study. Performance trends are analysed and the tuned population size for PSO across dimensi... Read More about Unexplained Fluctuations in Particle Swarm Optimisation Performance with Increasing Problem Dimensionality.

From Fitness Landscapes to Explainable AI and Back (2023)
Presentation / Conference Contribution
Thomson, S. L., Adair, J., Brownlee, A. E. I., & van den Berg, D. (2023, July). From Fitness Landscapes to Explainable AI and Back. Presented at GECCO '23, Lisbon, Portugal

We consider and discuss the ways in which search landscapes might contribute to the future of explainable artificial intelligence (XAI), and vice versa. Landscapes are typically used to gain insight into algorithm search dynamics on optimisation prob... Read More about From Fitness Landscapes to Explainable AI and Back.

Randomness in Local Optima Network Sampling (2023)
Presentation / Conference Contribution
Thomson, S. L., Veerapen, N., Ochoa, G., & van den Berg, D. (2023, July). Randomness in Local Optima Network Sampling. Presented at GECCO '23, Lisbon, Portugal

We consider statistical randomness in the construction of local optima networks (LONs) and conduct a preliminary and exploratory study into this. LONs capture a fitness landscape into network format: the nodes are local optima, and edges are heuristi... Read More about Randomness in Local Optima Network Sampling.

Channel Configuration for Neural Architecture: Insights from the Search Space (2023)
Presentation / Conference Contribution
Thomson, S. L., Ochoa, G., Veerapen, N., & Michalak, K. (2023, July). Channel Configuration for Neural Architecture: Insights from the Search Space. Presented at GECCO '23, Lisbon, Portugal

We consider search spaces associated with neural network channel configuration. Architectures and their accuracy are visualised using low-dimensional Euclidean embedding (LDEE). Optimisation dynamics are captured using local optima networks (LONs). L... Read More about Channel Configuration for Neural Architecture: Insights from the Search Space.

Frequency Fitness Assignment on JSSP: A Critical Review (2023)
Presentation / Conference Contribution
de Bruin, E., Thomson, S. L., & Berg, D. V. D. (2023, April). Frequency Fitness Assignment on JSSP: A Critical Review. Presented at EvoApplications 2023: Applications of Evolutionary Computation, Brno, Czech Republic

Metaheuristic navigation towards rare objective values instead of good objective values: is it a good idea? We will discuss the closed and open ends after presenting a successful replication study of Weise et al.’s ‘frequency fitness assignment’ for... Read More about Frequency Fitness Assignment on JSSP: A Critical Review.

Universally Hard Hamiltonian Cycle Problem Instances (2022)
Presentation / Conference Contribution
Sleegers, J., Thomson, S. L., & van den Berg, D. (2022, November). Universally Hard Hamiltonian Cycle Problem Instances. Presented at ECTA 2022 : 14th International Conference on Evolutionary Computation Theory and Applications, Valletta, Malta

In 2021, evolutionary algorithms found the hardest-known yes and no instances for the Hamiltonian cycle problem. These instances, which show regularity patterns, require a very high number of recursions for the best exact backtracking algorithm (Vand... Read More about Universally Hard Hamiltonian Cycle Problem Instances.