Keep Learning

Stalling in Space: Attractor Analysis for any Algorithm (2025)
Presentation / Conference Contribution
Thomson, S. L., Renau, Q., Vermetten, D., Hart, E., van Stein, N., & Kononova, A. V. (2025, April). Stalling in Space: Attractor Analysis for any Algorithm. Paper presented at EvoStar 2025, Trieste, Italy

Network-based representations of fitness landscapes have grown in popularity in the past decade; this is probably because of growing interest in explainability for optimisation algorithms. Local optima networks (LONs) have been especially dominant in... Read More about Stalling in Space: Attractor Analysis for any Algorithm.

Algorithm Selection with Probing Trajectories: Benchmarking the Choice of Classifier Model (2025)
Presentation / Conference Contribution
Renau, Q., & Hart, E. (2025, April). Algorithm Selection with Probing Trajectories: Benchmarking the Choice of Classifier Model. Presented at EvoSTAR 2025, Trieste, Italy

Recent approaches to training algorithm selectors in the black-box optimisation domain have advocated for the use of training data that is 'algorithm-centric' in order to encapsulate information about how an algorithm performs on an instance, rather... Read More about Algorithm Selection with Probing Trajectories: Benchmarking the Choice of Classifier Model.

Beyond the Hype: Benchmarking LLM-Evolved Heuristics for Bin Packing (2025)
Presentation / Conference Contribution
Sim, K., Hart, E., & Renau, Q. (2025, April). Beyond the Hype: Benchmarking LLM-Evolved Heuristics for Bin Packing. Presented at EvoSTAR 2025, Trieste, Italy

Coupling Large Language Models (LLMs) with Evolutionary Algorithms has recently shown significant promise as a technique to design new heuristics that outperform existing methods, particularly in the field of combinatorial optimisation. An escalating... Read More about Beyond the Hype: Benchmarking LLM-Evolved Heuristics for Bin Packing.

An Evaluation of Domain-agnostic Representations to Enable Multi-task Learning in Combinatorial Optimisation (2025)
Presentation / Conference Contribution
Stone, C., Renau, Q., Miguel, I., & Hart, E. (2024, June). An Evaluation of Domain-agnostic Representations to Enable Multi-task Learning in Combinatorial Optimisation. Presented at 18th Learning and Intelligent Optimization Conference, Ischia, Italy

We address the question of multi-task algorithm selection in combinatorial optimisation domains. This is motivated by a desire to simplify the algorithm-selection pipeline by developing a more general classifier that does not require specialised info... Read More about An Evaluation of Domain-agnostic Representations to Enable Multi-task Learning in Combinatorial Optimisation.

Identifying Easy Instances to Improve Efficiency of ML Pipelines for Algorithm-Selection (2024)
Presentation / Conference Contribution
Renau, Q., & Hart, E. (2024, September). Identifying Easy Instances to Improve Efficiency of ML Pipelines for Algorithm-Selection. Presented at 18th International Conference, PPSN 2024, Hagenberg, Austria

Algorithm-selection (AS) methods are essential in order to obtain the best performance from a portfolio of solvers over large sets of instances. However, many AS methods rely on an analysis phase, e.g. where features are computed by sampling solution... Read More about Identifying Easy Instances to Improve Efficiency of ML Pipelines for Algorithm-Selection.

Evaluating the Robustness of Deep-Learning Algorithm-Selection Models by Evolving Adversarial Instances (2024)
Presentation / Conference Contribution
Hart, E., Renau, Q., Sim, K., & Alissa, M. (2024, September). Evaluating the Robustness of Deep-Learning Algorithm-Selection Models by Evolving Adversarial Instances. Presented at 18th International Conference on Parallel Problem Solving From Nature PPSN 2024, Hagenburg, Austria

Deep neural networks (DNN) are increasingly being used to perform algorithm-selection in combinatorial optimisation domains, particularly as they accommodate input representations which avoid designing and calculating features. Mounting evidence fro... Read More about Evaluating the Robustness of Deep-Learning Algorithm-Selection Models by Evolving Adversarial Instances.

Improving Algorithm-Selectors and Performance-Predictors via Learning Discriminating Training Samples (2024)
Presentation / Conference Contribution
Renau, Q., & Hart, E. (2024, July). Improving Algorithm-Selectors and Performance-Predictors via Learning Discriminating Training Samples. Presented at GECCO 2024, Melbourne, Australia

The choice of input-data used to train algorithm-selection models is recognised as being a critical part of the model success. Recently, feature-free methods for algorithm-selection that use short trajec-tories obtained from running a solver as input... Read More about Improving Algorithm-Selectors and Performance-Predictors via Learning Discriminating Training Samples.

Learning Descriptors for Novelty-Search Based Instance Generation via Meta-evolution (2024)
Presentation / Conference Contribution
Marrero, A., Segredo, E., León, C., & Hart, E. (2024, July). Learning Descriptors for Novelty-Search Based Instance Generation via Meta-evolution. Presented at GECCO '24: Genetic and Evolutionary Computation Conference, Melbourne, Australia

The ability to generate example instances from a domain is important in order to benchmark algorithms and to generate data that covers an instance-space in order to train machine-learning models for algorithm selection. Quality-Diversity (QD) algorit... Read More about Learning Descriptors for Novelty-Search Based Instance Generation via Meta-evolution.

Synthesising Diverse and Discriminatory Sets of Instances using Novelty Search in Combinatorial Domains (2024)
Journal Article
Marrero, A., Segredo, E., Leon, C., & Hart, E. (online). Synthesising Diverse and Discriminatory Sets of Instances using Novelty Search in Combinatorial Domains. Evolutionary Computation, https://doi.org/10.1162/evco_a_00350

Gathering sufficient instance data to either train algorithm-selection models or understand algorithm footprints within an instance space can be challenging. We propose an approach to generating synthetic instances that are tailored to perform well w... Read More about Synthesising Diverse and Discriminatory Sets of Instances using Novelty Search in Combinatorial Domains.

On the Utility of Probing Trajectories for Algorithm-Selection (2024)
Presentation / Conference Contribution
Renau, Q., & Hart, E. (2024, April). On the Utility of Probing Trajectories for Algorithm-Selection. Presented at EvoStar 2024, Aberystwyth, UK

Machine-learning approaches to algorithm-selection typically take data describing an instance as input. Input data can take the form of features derived from the instance description or fitness landscape , or can be a direct representation of the ins... Read More about On the Utility of Probing Trajectories for Algorithm-Selection.