Dennis G. Wilson
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Dennis G. Wilson has a total of 60 co-authors. Here are the top co-authors by number of shared publications:
- Sylvain Cussat-Blanc
Institut de Recherche en Informatique de Toulouse (IRIT) , Toulouse 31400
13 shared publications - Camilo De La Torre
Université Toulouse Capitole, IRIT - CNRS UMR5505
10 shared publications - Giorgia Nadizar
University of Toulouse Capitole, Toulouse, France
8 shared publications - Hervé Luga
Université Toulouse Capitole, IRIT - CNRS UMR5505
8 shared publications - Yuri Lavinas
University Toulouse Capitole, IRIT - CNRS UMR5505
7 shared publications - Rafael Almar
LEGOS (Université de Toulouse/CNRS/IRD/CNES) Toulouse France
6 shared publications
Latest Publications
- 25 August 2026
- Published by Springer Nature
- p. 233-250
Structure vs. Constants: A Study of Coefficients in Cartesian Genetic Programming
- 25 August 2026
- Published by Springer Nature
- p. 233-250
Structure vs. Constants: A Study of Coefficients in Cartesian Genetic Programming
- Open Access
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1688-1704
Cartesian Genetic Programming - From foundations to recent developments and applications
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1688-1704
Open AccessCartesian Genetic Programming - From foundations to recent developments and applications
- 3 June 2026
- Published by Springer Nature
- p. 79-101
On Interpretability in Multimodal Biomedical Image Analysis
- 3 June 2026
- Published by Springer Nature
- p. 79-101
On Interpretability in Multimodal Biomedical Image Analysis
- Open Access
- 28 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
Restart of the JEDi: Dynamic Quality with Just Enough Diversity
In complex control tasks, finding high-performing policies often requires discovering and exploiting specific behavioral strategies. While Quality-Diversity (QD) algorithms can uncover these strategies through extensive behavior space exploration, they sacrifice efficiency by improving suboptimal behaviors. Conversely, Evolution Strategies (ES) achieve impressive performance through focused optimization but frequently become trapped in local optima due to limited behavioral exploration. We present Quality with J ust E nough Di versity (JEDi), a new optimization framework that resolves this fundamental tension. JEDi employs a combination of Gaussian Process modeling and parallel Evolution Strategies to intelligently explore behavioral space while maintaining focused optimization. At its core, JEDi learns a probabilistic mapping between behaviors and performance, using this model to identify and target promising behavioral regions that could unlock better solutions. This targeted exploration is achieved through multiple Evolution Strategy emitters that simultaneously optimize toward selected behaviors while maximizing task performance. To further improve JEDi’s exploration capabilities, we introduce its Dy namic variant DyJEDi with an adaptive restart mechanism that dynamically detects and responds to emitter convergence, independently restarting each Evolution Strategy when it stagnates in both behavior and fitness space. This dynamic approach significantly improves exploration efficiency and robustness to local optima. We demonstrate that DyJEDi outperforms both traditional Evolution Strategies and Quality-Diversity approaches across challenging continuous robotics control tasks, achieving higher final performance. Most notably, DyJEDi solves several hard exploration problems where standard ES methods consistently fail.
- 28 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
Open AccessRestart of the JEDi: Dynamic Quality with Just Enough Diversity
In complex control tasks, finding high-performing policies often requires discovering and exploiting specific behavioral strategies. While Quality-Diversity (QD) algorithms can uncover these strategies through extensive behavior space exploration, they sacrifice efficiency by improving suboptimal behaviors. Conversely, Evolution Strategies (ES) achieve impressive performance through focused optimization but frequently become trapped in local optima due to limited behavioral exploration. We present Quality with J ust E nough Di versity (JEDi), a new optimization framework that resolves this fundamental tension. JEDi employs a combination of Gaussian Process modeling and parallel Evolution Strategies to intelligently explore behavioral space while maintaining focused optimization. At its core, JEDi learns a probabilistic mapping between behaviors and performance, using this model to identify and target promising behavioral regions that could unlock better solutions. This targeted exploration is achieved through multiple Evolution Strategy emitters that simultaneously optimize toward selected behaviors while maximizing task performance. To further improve JEDi’s exploration capabilities, we introduce its Dy namic variant DyJEDi with an adaptive restart mechanism that dynamically detects and responds to emitter convergence, independently restarting each Evolution Strategy when it stagnates in both behavior and fitness space. This dynamic approach significantly improves exploration efficiency and robustness to local optima. We demonstrate that DyJEDi outperforms both traditional Evolution Strategies and Quality-Diversity approaches across challenging continuous robotics control tasks, achieving higher final performance. Most notably, DyJEDi solves several hard exploration problems where standard ES methods consistently fail.
- 9 May 2026
- Published by Springer Nature
- p. 501-517
Multi-objective Optimization for Synthetic-to-Real Style Transfer
- 9 May 2026
- Published by Springer Nature
- p. 501-517
Multi-objective Optimization for Synthetic-to-Real Style Transfer
- 28 April 2026
- Published by Springer Nature
- p. 34-50
A Comparative Study on Robustness in Evolved Image Classifiers
- 28 April 2026
- Published by Springer Nature
- p. 34-50
A Comparative Study on Robustness in Evolved Image Classifiers
- Open Access
- 11 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
Analyzing Interpretable Visual Control Policy Search and Synthesis
Vision-based decision-making is relevant to many domains, including safety-critical ones where transparency matters as much as performance. Therefore, automating sequential decision making in such settings requires approaches that balance effectiveness with interpretability. While deep reinforcement learning techniques based on artificial neural networks have achieved strong performance, their black-box nature typically necessitates post hoc explainability analyses. To address this limitation, we propose an approach based on Graph-based Genetic Programming (GGP) that generates policies in the form of computer code, which is fully observable and inherently interpretable. To improve both performance and robustness, we expose GGP-based visual control policies to multiple representative conditions during optimization, mitigating convergence to strong local optima and fragility under changing conditions. Finally, to gain insights into the evolutionary search dynamics, we employ Search Trajectory Networks, an analytical and visualization tool for studying optimization behavior. Our results demonstrate that the resulting policies approach human-level performance and empirically confirm the presence of strong local optima acting as attractors during evolution, providing new insights into the behavior and potential limitations of interpretable evolutionary policy search approaches.
- 11 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
Open AccessAnalyzing Interpretable Visual Control Policy Search and Synthesis
Vision-based decision-making is relevant to many domains, including safety-critical ones where transparency matters as much as performance. Therefore, automating sequential decision making in such settings requires approaches that balance effectiveness with interpretability. While deep reinforcement learning techniques based on artificial neural networks have achieved strong performance, their black-box nature typically necessitates post hoc explainability analyses. To address this limitation, we propose an approach based on Graph-based Genetic Programming (GGP) that generates policies in the form of computer code, which is fully observable and inherently interpretable. To improve both performance and robustness, we expose GGP-based visual control policies to multiple representative conditions during optimization, mitigating convergence to strong local optima and fragility under changing conditions. Finally, to gain insights into the evolutionary search dynamics, we employ Search Trajectory Networks, an analytical and visualization tool for studying optimization behavior. Our results demonstrate that the resulting policies approach human-level performance and empirically confirm the presence of strong local optima acting as attractors during evolution, providing new insights into the behavior and potential limitations of interpretable evolutionary policy search approaches.
- 4 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Evolving principes of artificial neural design
Évolution des principes de la conception des réseaux de neurones artificiels: Le cerveau biologique est composé d'un ensemble d'éléments qui évoluent depuis des millions d'années. Les neurones et autres cellules forment un réseau complexe d'interactions duquel émerge l'intelligence. Bon nombre de concepts neuronaux provenant de l’étude du cerveau biologique ont été utilisés dans des modèles informatiques pour développer les algorithmes d’intelligence artificielle. C'est particulièrement le cas des réseaux neuronaux profonds modernes qui révolutionnent actuellement de nombreux domaines de recherche en informatique tel que la vision par ordinateur, la traduction automatique, le traitement du langage naturel et bien d'autres. Cependant, les réseaux de neurones artificiels ne sont basés que sur un petit sous-ensemble de fonctionnalités biologiques du cerveau. Ils se concentrent souvent sur les fonctions globales, homogènes et à un système complexe et localement hétérogène. Dans cette thèse, nous avons d'examiner le cerveau biologique, des neurones simples aux réseaux capables d'apprendre. Nous avons examiné individuellement la cellule neuronale, la formation des connexions entre les cellules et comment un réseau apprend au fil du temps. Pour chaque composant, nous avons utilisé l'évolution artificielle pour trouver les principes de conception neuronale qui nous avons optimisés pour les réseaux neuronaux artificiels. Nous proposons aussi un modèle fonctionnel du cerveau qui peut être utilisé pour étudier plus en profondeur certains composants du cerveau, incluant toutes les fonctions conçues pour l'optimisation automatique telles que l'évolution. Notre objectif est d'améliorer la performance des réseaux de neurones artificiels par les moyens inspirés des neurosciences modernes. Cependant, en évaluant les effets biologiques dans le contexte d'un agent virtuel, nous espérons également fournir des modèles de cerveau utiles aux biologistes.
- 4 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Evolving principes of artificial neural design
Évolution des principes de la conception des réseaux de neurones artificiels: Le cerveau biologique est composé d'un ensemble d'éléments qui évoluent depuis des millions d'années. Les neurones et autres cellules forment un réseau complexe d'interactions duquel émerge l'intelligence. Bon nombre de concepts neuronaux provenant de l’étude du cerveau biologique ont été utilisés dans des modèles informatiques pour développer les algorithmes d’intelligence artificielle. C'est particulièrement le cas des réseaux neuronaux profonds modernes qui révolutionnent actuellement de nombreux domaines de recherche en informatique tel que la vision par ordinateur, la traduction automatique, le traitement du langage naturel et bien d'autres. Cependant, les réseaux de neurones artificiels ne sont basés que sur un petit sous-ensemble de fonctionnalités biologiques du cerveau. Ils se concentrent souvent sur les fonctions globales, homogènes et à un système complexe et localement hétérogène. Dans cette thèse, nous avons d'examiner le cerveau biologique, des neurones simples aux réseaux capables d'apprendre. Nous avons examiné individuellement la cellule neuronale, la formation des connexions entre les cellules et comment un réseau apprend au fil du temps. Pour chaque composant, nous avons utilisé l'évolution artificielle pour trouver les principes de conception neuronale qui nous avons optimisés pour les réseaux neuronaux artificiels. Nous proposons aussi un modèle fonctionnel du cerveau qui peut être utilisé pour étudier plus en profondeur certains composants du cerveau, incluant toutes les fonctions conçues pour l'optimisation automatique telles que l'évolution. Notre objectif est d'améliorer la performance des réseaux de neurones artificiels par les moyens inspirés des neurosciences modernes. Cependant, en évaluant les effets biologiques dans le contexte d'un agent virtuel, nous espérons également fournir des modèles de cerveau utiles aux biologistes.
- Open Access
- 28 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Interpretable machine learning for shoreline forecasting
Machine learning has revolutionized scientific modeling, providing breakthroughs in fields ranging from weather prediction to protein folding. However, its adoption in physics-based domains remains limited due to the lack of interpretability in traditional black-box models. In environmental sciences, where understanding underlying mechanisms is critical, symbolic regression offers an alternative by discovering transparent mathematical expressions that can represent physical principles. In this work, we demonstrate the application of symbolic regression to physical modeling through shoreline prediction, a critical area for understanding coastal evolution under climate change and human influence. Unlike traditional physics-based models, which rely on assumptions that may not generalize across diverse coastal environments, our approach evolves a population of interpretable models directly from global observational data. By optimizing both predictive accuracy and model complexity, we uncover region-specific formulations that reveal the dominant physical drivers of shoreline change. This methodology enables data-driven discovery while maintaining alignment with physical intuition, providing new insights into physical dynamics across multiple spatial and temporal scales.
- 28 February 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessInterpretable machine learning for shoreline forecasting
Machine learning has revolutionized scientific modeling, providing breakthroughs in fields ranging from weather prediction to protein folding. However, its adoption in physics-based domains remains limited due to the lack of interpretability in traditional black-box models. In environmental sciences, where understanding underlying mechanisms is critical, symbolic regression offers an alternative by discovering transparent mathematical expressions that can represent physical principles. In this work, we demonstrate the application of symbolic regression to physical modeling through shoreline prediction, a critical area for understanding coastal evolution under climate change and human influence. Unlike traditional physics-based models, which rely on assumptions that may not generalize across diverse coastal environments, our approach evolves a population of interpretable models directly from global observational data. By optimizing both predictive accuracy and model complexity, we uncover region-specific formulations that reveal the dominant physical drivers of shoreline change. This methodology enables data-driven discovery while maintaining alignment with physical intuition, providing new insights into physical dynamics across multiple spatial and temporal scales.
- Open Access
Multi-Objective Optimization for Synthetic-to-Real Style Transfer
Semantic segmentation networks require large amounts of pixel-level annotated data, which are costly to obtain for real-world images. Computer graphics engines can generate synthetic images alongside their ground-truth annotations. However, models trained on such images can perform poorly on real images due to the domain gap between real and synthetic images. Style transfer methods can reduce this difference by applying a realistic style to synthetic images. Choosing effective data transformations and their sequence is difficult due to the large combinatorial search space of style transfer operators. Using multi-objective genetic algorithms, we optimize pipelines to balance structural coherence and style similarity to target domains. We study the use of paired-image metrics on individual image samples during evolution to enable rapid pipeline evaluation, as opposed to standard distributional metrics that require the generation of many images. After optimization, we evaluate the resulting Pareto front using distributional metrics and segmentation performance. We apply this approach to standard datasets in synthetic-to-real domain adaptation: from the video game GTA5 to real image datasets Cityscapes and ACDC, focusing on adverse conditions. Results demonstrate that evolutionary algorithms can propose diverse augmentation pipelines adapted to different objectives. The contribution of this work is the formulation of style transfer as a sequencing problem suitable for evolutionary optimization and the study of efficient metrics that enable feasible search in this space. The source code is available at: https://github.com/echigot/MOOSS.
Open AccessMulti-Objective Optimization for Synthetic-to-Real Style Transfer
Semantic segmentation networks require large amounts of pixel-level annotated data, which are costly to obtain for real-world images. Computer graphics engines can generate synthetic images alongside their ground-truth annotations. However, models trained on such images can perform poorly on real images due to the domain gap between real and synthetic images. Style transfer methods can reduce this difference by applying a realistic style to synthetic images. Choosing effective data transformations and their sequence is difficult due to the large combinatorial search space of style transfer operators. Using multi-objective genetic algorithms, we optimize pipelines to balance structural coherence and style similarity to target domains. We study the use of paired-image metrics on individual image samples during evolution to enable rapid pipeline evaluation, as opposed to standard distributional metrics that require the generation of many images. After optimization, we evaluate the resulting Pareto front using distributional metrics and segmentation performance. We apply this approach to standard datasets in synthetic-to-real domain adaptation: from the video game GTA5 to real image datasets Cityscapes and ACDC, focusing on adverse conditions. Results demonstrate that evolutionary algorithms can propose diverse augmentation pipelines adapted to different objectives. The contribution of this work is the formulation of style transfer as a sequencing problem suitable for evolutionary optimization and the study of efficient metrics that enable feasible search in this space. The source code is available at: https://github.com/echigot/MOOSS.