Evolutionary computation (EC) has long been at the forefront of harnessing the principles of natural selection to solve complex problems in a myriad of domains, ranging from optimization to artificial intelligence. Recently, the fusion of EC with machine learning (ML) techniques has opened new vistas for research, offering unprecedented opportunities to tackle some of the most challenging issues faced by the scientific community today. This intersectional domain leverages the explorative power of evolutionary algorithms to enhance the adaptability and efficiency of machine learning models, leading to breakthroughs in algorithmic design, problem-solving methodologies, and the interpretability of artificial intelligence systems.
The Toulouse Workshop on Evolutionary Computation and Machine Learning will explore the interactions between EC and ML. This workshop will spotlight a diverse array of topics, including but not limited to the evolutionary design of explainable algorithms for biomedical image segmentation, the deployment of graph genetic programming for interpretable control, advancements in active learning, and the quest for quality and diversity in reinforcement learning. With an international panel of speakers converging in Toulouse, this event will be a platform for showcasing how evolutionary principles can drive the creation of diverse forms of artificial intelligence.
Details
April 23rd 2024
9am - 6pm
Université Toulouse Capitole
Manufacture des Tabacs
21 All. de Brienne, 31015 Toulouse
Program
| Time | Speaker | Title |
|---|---|---|
| 9:00 | coffee | |
| 9:30 | Antoine Cully | Adaptive Machines: Quality-Diversity and Evolutionary-RL for Versatile and Resilient Robotics |
| 10:15 | Manon Flageat | Consistent diversity: importance of reproducible solutions when optimizing for diversity |
| 10:45 | Paul-Antoine le Tolguenec | Curiosity creates Diversity in Policy Search |
| 11:15 | Giorgia Nadizar | Interpretable Control with Graph Genetic Programming |
| 12:00 | lunch | |
| 13:30 | Eric Medvet | Where is intelligence in simulated modular soft robots? |
| 14:00 | Sylvain Cussat-Blanc | Evolutionary design of explainable algorithms for biomedical image segmentation |
| 14:45 | Yuri Lavinas | Active Learning in Genetic Programming |
| 15:15 | coffee | |
| 15:45 | Gabriela Ochoa | Search Trajectory Networks |
| 16:15 | Camilo de la Torre | Search Trajectory Networks for Cartesian Genetic Programming |
| 16:30 | Tarek Kunze | Searching Search Spaces: Meta-evolving a Geometric Encoding for Neural Networks |
| 16:45 | Dennis Wilson | Evolving Programs with Large Language Models |
Registration
In-person registration will be limited to 40 participants. To register, please fill out this form.
Online registration is appreciated but not necessary. Online participation will be available through Zoom:
https://zoom.us/my/dennisgwilson
Paul Templier thesis defense
The TEML workshop follows the PhD defense of Paul Templier, entitled Leveraging Structures in Evolutionary Neural Policy Search.
The defense will take place on Monday 22 April at 2:00pm in the salle des thèses of ISAE-SUPAERO, and will be conducted in English.
A webcast will be also available on Zoom.
The jury will be composed of:
- Emmanuel RACHELSON, ISAE-SUPAERO, PhD supervisor
- Dennis WILSON, ISAE-SUPAERO, PhD co-supervisor
- Nikolaus HANSEN, INRIA, Reviewer
- Olivier SIGAUD, Sorbonne Université, Reviewer
- Gabriela OCHOA, University of Stirling, Examiner
- Daniel DELAHAYE, ENAC, Examiner
Abstract:
While training an artificial agent for complex tasks like driving a car, mastering a video game, or controlling plasma in a nuclear fusion reactor, innovations can lead to intelligent behavior. In such scenarios, a promising approach is to mimick the natural world’s evolutionary process, which has honed the problem-solving capabilities of animal brains. Evolutionary Neural Policy Search (ENPS) draws inspiration from this concept. It creates a diverse population of “brains” represented by neural networks, allowing the system to “evolve” by selectively combining and mutating successful individuals. This thesis delves into the core components of ENPS and their intricate interplay. By analyzing the structures of ENPS, the goal is to design novel policy search methods that enhance these components, ultimately leading to the development of more efficient and effective learning algorithms for complex tasks.
Workshop Organizers
Dennis G. Wilson
email
ISAE-Supaero, University of Toulouse, France
https://pagespro.isae-supaero.fr/dennis-wilson
Sylvain Cussat-Blanc
email
IRIT, University of Toulouse, France
https://www.irit.fr/~Sylvain.Cussat-Blanc/