Volume 18, Issue 1

Newsletter of the ACM Special Interest Group on Genetic and Evolutionary Computation

Editorial

Welcome to the 1st (Spring) 2025 issue of the SIGEvolution newsletter! We start with an overview of the 2024 Evolutionary Computation Theory and Applications (ECTA) conference, a yearly event held in Porto, Portugal last year. We follow with reports of two excellent PhD dissertations completed in 2024. Our cover image relates to one of them, showing a mosaic of optimisation landscapes, depicted as constraint violation values, with feasible solutions shown in white. We continue with the the latest issues of the main journals in evolutionary computation, forthcoming events, and calls for papers. As ever, please get in touch if you have contributions or suggestions for the newsletter.

Gabriela Ochoa, Editor.

Conference Overview

Evolutionary Computation Theory and Applications (ECTA) 2024

Jacob de Nobel, Leiden University, The Netherlands

Last year’s Evolutionary Computation Theory and Applications (ECTA 2024) conference took place in the beautiful city of Porto, Portugal, from November 20th to 22nd, 2024, as part of the International Joint Conference on Computational Intelligence (IJCCI). The event, organized by INSTICC and supported by ACM SIGAI, AAAI, and other esteemed institutions, brought together researchers worldwide to discuss their latest advancements.

The conference began with a welcome session, where attendees were greeted by program chairs Francesco Marcelloni, Kurosh Madani, Niki van Stein, and conference chair Joaquim Filipe.

This was followed by an insightful keynote by Tome Eftimov from the Jožef Stefan Institute, who spoke on “Trustworthy Benchmarking for Black-Box Single-Objective Optimization.” His talk highlighted recent advancements in benchmarking and the importance of moving beyond simple performance improvements, towards better understanding algorithm behavior and problem characteristics.

Presentations and discussions covered various topics throughout the first day, from Evolutionary Multi-Objective Optimization to Neural Computation Theory. A key advantage of ECTA being part of IJCCI is the opportunity for participants to explore related fields. Attendees could easily join sessions from the Neural Computation Theory and Applications (NCTA) and Fuzzy Computation Theory and Applications (FCTA) subconferences, gaining a broader perspective on computational intelligence.

This year, EXPLAINS, a new track focused on Explainable AI, was also introduced, which generated significant interest. Talks in this session explored ways to make AI models more transparent and interpretable. Among the notable presentations were the ones from my colleagues at LIACS: Sofoklis Kitharidis’ discussion on Fair Machine Learning at NCTA and Qi Huang’s talk on TX-Gen, a novel method for generating counterfactual examples for time-series classification at EXPLAINS.
Wednesday evening’s poster session provided a great opportunity for informal discussions over coffee, tea, and traditional Portuguese treats like Pastel de Nata. Researchers engaged in lively conversations while showcasing their work in a more interactive setting.

Thursday began with a keynote by Rita Ribeiro, who presented her research on predictive maintenance for Industry 4.0 and 5.0, emphasizing the importance of bridging theoretical advancements with real-world applications. One of the standout sessions of the day was on Behavior Analysis of Evolutionary Algorithms, where I had the opportunity to present my work on “Sampling in CMA-ES: Low Numbers of Low Discrepancy Points.” I was honored to receive the Best Paper Award for ECTA 2024, which made the trip even more worthwhile.

Later that evening, attendees enjoyed a guided visit to “Caves Taylor,” one of Porto’s historic port wine cellars. The tour provided insight into the craftsmanship behind port wine, followed by a dinner accompanied by traditional Portuguese music. The dinner was amazing, with spectacular views from the restaurant.

Friday’s program featured a compelling keynote by Gabriela Ochoa from the University of Stirling titled “Search Trajectories Illuminated.” Her talk explored search behaviors in optimization problems, providing valuable insights into new ways of looking at optimization problems from a graph theory perspective.

ECTA 2024 was a productive and engaging conference, fostering meaningful discussions and collaborations. Reflecting on the experience, I am grateful to the organizers, speakers, and fellow researchers who contributed to its success.

Looking forward to continuing these discussions next year during ECTA 2025 in Marbella, Spain!

About the Author

Jacob de Nobel (left in the photo) is a PhD student at LIACS, Leiden University, and a core developer for IOHprofiler. His research focuses on applying and evaluating optimization algorithms, particularly in developing improved speech encoding strategies for cochlear implants—neuroprosthetic devices for individuals with profound hearing loss. He is especially interested in understanding the behavior of continuous optimization algorithms, assessing their performance, and refining benchmarking techniques to ensure their effectiveness in practical applications

PhD Dissertation Reports

Characterization of Constrained Continuous Multiobjective Optimization Problems

Aljoša Vodopija, Jožef Stefan Institute, Ljubljana, Slovenia

Supervisor: Prof. Bogdan Filipič, Jožef Stefan Institute, Ljubljana, Slovenia

Constrained multiobjective optimization problems (CMOPs) are integral to many real-world applications. Despite significant recent advances in multiobjective optimization, there remains a lack of adequate characterization and understanding of CMOPs, which limits the development and benchmarking of optimizers. This thesis addresses this gap by investigating CMOPs from two complementary perspectives: the feature space and the performance space.

Concerning the feature space, the research extends traditional exploratory landscape analysis techniques to CMOPs, proposing novel features to characterize CMOP landscapes. Special attention is given to the multimodality of constraint violations (see Figure 1, which depicts problem landscapes of three well-known CMOPs), allowing for a deeper understanding of problems and their impact on optimizer performance. By comparing artificial test problems with real-world problems, the analysis reveals that artificial test problems often fail to replicate the complexity of real-world problems, such as strong negative correlations between objectives and constraints.

Figure 1: Problem landscapes of three well-known CMOPs (C2-DTLZ2, MW6, and DAS-CMOP1) depicted as constraint violation values, with feasible solutions shown in white.

Regarding the performance space, a new methodology is introduced to evaluate optimizer performance in solving CMOPs, simultaneously addressing the goals of Pareto front approximation and constraint satisfaction. This approach is then used to assess how well different test problems distinguish optimizers’ performance, providing insights into their effectiveness for benchmarking purposes. The results indicate that most of the complexity in solving artificial test problems arises from approximating the Pareto front, while constraint satisfaction is often trivial to achieve.

The proposed methodology is applied to two real-world problems. In the design of cyclone dust separators, newly proposed landscape features characterize the problem, highlighting the challenges faced by existing optimizers in effectively handling constraints. The second problem, elevator group control, is a more complex problem that poses greater challenges in terms of solution diversity and convergence. The study underscores the necessity of combining the feature and performance space analyses to address these problems efficiently.

Here is a link to the full dissertation: https://shorturl.at/K91fH

Related Publications

  • Vodopija, A., Tušar, T., & Filipič, B. (2025). Characterization of constrained continuous multiobjective optimization problems: A performance space perspective. IEEE Trans. Evol. Comput., 29(1), 275–285. https://doi.org/10.1109/TEVC.2024.3366659
  • Vodopija, A., Tušar, T., & Filipič, B. (2022). Characterization of constrained continuous multiobjective optimization problems: A feature space perspective. Inf. Sci., 607, 244–262. https://doi.org/10.1016/j.ins.2022.05.106
  • Vodopija, A., Stork, J., Bartz-Beielstein, T., & Filipič, B. (2022). Elevator group control as a constrained multiobjective optimization problem. Appl. Soft Comput., 115, 108277. https://doi.org/10.1016/j.asoc.2021.108277
  • Vodopija, A., Breiderhoff, B., Naujoks, B., & Filipič, B. (2021). Design of cyclone dust separators: A constrained multiobjective optimization perspective. 2021 IEEE Congress on Evolutionary Computation (CEC), 1983–1990. https://doi.org/10.1109/CEC45853.2021.9504991
  • Vodopija, A., Tušar, T., & Filipič, B. (2021). Analyzing the diversity of constrained multiobjective optimization test suites. Slovenian Conference on Artificial Intelligence: Proceedings of the 24th International Multiconference (IS 2021), 51–54.
  • Vodopija, A., Janko, V., Luštrek, M., & Filipič, B. (2020). Constrained multiobjective optimization for the design of energy-efficient context recognition systems. Bioinspired Optimization Methods and Their Applications: 9th International Conference (BIOMA 2020), 308–320. https://doi.org/10.1007/978-3-030-63710-1_24
  • Vodopija, A., Oyama, A., & Filipič, B. (2019). Ensemble-based constraint handling in multiobjective optimization. Proceedings of the Genetic and Evolutionary Computation Conference Companion (GECCO ’19), 2072–2075. https://doi.org/10.1145/3319619.3326909
  • Vodopija, A., Stork, J., Bartz-Beielstein, T., & Filipič, B. (2018). Model-based multiobjective optimization of elevator group control. International Conference on High-Performance Optimization in Industry: Proceedings of the 21st International Multiconference Information Society (IS 2018), 43–46.

About the Author

Aljoša Vodopija is a research assistant at the Department of Intelligent Systems of the Jožef Stefan Institute, Ljubljana, Slovenia, and a data scientist at Outbrain Inc., Ljubljana, Slovenia. In 2024, he earned his Ph.D. with the highest honors, Summa Cum Laude, from the Jožef Stefan International Postgraduate School for his thesis titled Characterization of Constrained Continuous Multiobjective Optimization Problems. His research interests include evolutionary computation and recommendation systems.

Leveraging Structures in Evolutionary Neural Policy Search

Paul Templier, ISAE-Supaero, University of Toulouse, France

Supervisors: Emmanuel Rachelson and Dennis Wilson, ISAE-Supaero, University of Toulouse, France

Training agents to perform complex tasks like driving a car, mastering a video game, or controlling a robot to walk presents a significant challenge when expert demonstrations are not available. In nature, complex behaviors and characteristics can emerge through evolution, as animals adapt to their environments and problems over generations.

Evolutionary Neural Policy Search replicates this natural process to discover effective policies for artificial agents, through three fundamental building blocks: neural networks that represent the agents (or “policies”), an environment that evaluates their performance, and an evolutionary algorithm to optimize them.

This thesis studies this framework, and more specifically Evolution Strategies (ES) in this context, to explore how a deeper understanding of these components and their interactions can lead to more effective design choices for learning methods. Throughout its exploration, this thesis builds bridges between ES and other domains of evolutionary computation like Quality-Diversity (QD) or Genetic Programming, but also with Reinforcement Learning (RL) and neural network representation. 

Evolutionary Neural Policy Search framework
Figure 1: Evolutionary Neural Policy Search framework

Neural Network Representation

First, optimizing neural networks with evolutionary methods requires to represent them as genomes, usually by concatenating all the weights and biases into one large vector. The size of that genome scales quadratically with the number of neurons, and large genomes can quickly become intractable for methods such as CMA-ES or xNES.

To be more efficient, we introduced the Geometric Encoding for NeuroEvolution (GENE), a novel representation that leverages the inherent structure of neural networks—organized as layers of interconnected neurons—to significantly shorten genomes. Rather than treating networks as unstructured parameter collections, GENE encodes information at the neuron level, creating a more efficient search space that still preserves performance while reducing optimization costs by 2 orders of magnitude.

We then refined GENE itself through meta-evolution, using Cartesian Genetic Programming to represent the encoding. This process naturally discovered encodings that favor very sparse network architectures, keeping only 3% of the connections, and corroborating recent research results in deep learning.

Genetic Drift Regularization

Interactions between a policy and an environment, such as picking actions to play in a game or sending instructions to a robot’s motors, often open the door to the application of Reinforcement Learning (RL) to learn a policy from all the individual steps.

Evolutionary methods, which learn from the episodic scores of solutions, are often combined with it by injecting the policy trained with RL, the “actor”, into the population.

When examining interactions between RL and ES specifically, we discovered through landscape analysis that the genomes of the actor and the ES tend to diverge, which can lead the injection of the actor to break the update of the ES. Genetic Drift Regularization (GDR) provides a straightforward yet effective solution to this problem, constraining the RL actor to remain genetically proximate to the ES distribution, and maintains the benefits of cross-paradigm integration while avoiding performance degradation.

Figure 2: MAP-Elites, ES and JEDi archives on a Walker2D task. The first row presents the score of the best policy for each behavior, the second row is the number of solutions evaluated.

Quality with Just Enough Diversity

Finally, Evolution Strategies are known to struggle on deceptive fitness landscapes, where estimating the fitness gradient will lead to suboptimal local optima. Leveraging behavior descriptors from Quality-Diversity, we introduced the framework of Quality with Just Enough Diversity (JEDi) that learns the relationship between behavior and fitness, to then focus on promising ways to solve the problem. JEDi solves the issues of ES on deceptive problems, but also reaches higher scores than QD methods on other tasks as they spend a lot of budget on exploration rather than exploitation.

References

About the Author

Paul Templier is a Research Associate at Imperial College London (UK), where he focuses on adaptive and intelligent robotics and on Quality-Diversity optimization. He obtained his PhD from ISAE-SUPAERO (Toulouse, France) in 2024, during which he researched methods to leverage structures of evolutionary policy search to improve learning mechanisms.

Journal Latest Issues

Evolutionary computation

Editors-in-Chief: Thomas Bäck and Hao Wang

March 2025

Articles

MO-SMAC: Multi-objective Sequential Model-based Algorithm Configuration
Jeroen G. Rook, Carolin Benjamins, Jakob Bossek, Heike Trautmann, Holger H. Hoos, Marius Lindauer

Solving Many-objective Optimization Problems based on PF Shape Classification and Vector Angle Selection. Y.T. Wu, F.Z. Ge, D.B. Chen, L. Shi

Beyond Landscape Analysis: DynamoRep Features For Capturing Algorithm-Problem Interaction In Single-Objective Continuous Optimization
Gjorgjina Cenikj, Gašper Petelin, Carola Doerr, Peter Korošec, Tome Eftimov

On the use of the Doubly Stochastic Matrix models for the Quadratic Assignment Problem
Valentino Santucci, Josu Ceberio

P-NP instance decomposition based on the Fourier transform for solving the Linear Ordering Problem
Xabier Benavides, Leticia Hernando, Josu Ceberio, Jose A. Lozano

Genetic Programming and Evolvable Machines (GPEM)

Editor-in-chief: Leonardo Trujillo

Editorial

Introduction for the special issue on highlights of genetic programming 2023 events

Gisele L. Pappa, Mario Giacobini, Domagoj Jakobović

Research Articles

Constraining genetic symbolic regression via semantic backpropagation
Maximilian Reissmann, Yuan Fang, Richard D. Sandberg

Memetic semantic boosting for symbolic regression
Alessandro Leite, Marc Schoenauer
Part of collection: Highlights of Genetic Programming 2023 Events

Book Reviews

“Machine learning assisted evolutionary multi- and many-objective optimization” by Dhish Kumar Saxena, Sukrit Mittal, Kalyanmoy Deb, and Erik D. Goodman. Springer, 2024
Saltuk Buğra Selçuklu

“The science of soft robots” by Koichi Suzumori, Kenjiro Fukuda, Ryuma Niiyama, and Kohei Nakajima. Springer, 2023
Eric MedvetErica Salvato

ACM Transactions on Evolutionary Learning and Optimization (TELO)

Editors-in-chief:  Jürgen Branke, Manuel López-Ibáñez

Volume 5, Issue 1. March 2025

Articles

Body and Brain Quality-Diversity in Robot Swarms
Sindiso Mkhatshwa, Geoff Nitschke

Synergizing Quality-Diversity with Descriptor-Conditioned Reinforcement Learning
Maxence Faldor,Félix Chalumeau, Manon Flageat, Antoine Cully

Covariance Matrix Adaptation MAP-Annealing: Theory and Experiments
Shihan Zhao, Bryon Tjanaka, Matthew C. Fontaine, Stefanos Nikolaidis

MA-BBOB: A Problem Generator for Black-Box Optimization Using Affine Combinations and Shifts
Diederick Vermetten, Furong Ye, Thomas Bäck, Carola Doerr

CMA-ES with Learning Rate Adaptation
Masahiro Nomura, Youhei Akimoto, Isao Ono

An Evolutionary Algorithm for Expensive Mixed-Integer Black-Box Optimization with Explicit Constraints
Yuan Hong, Dirk V. Arnold

Forthcoming Events

EvoStar 2025

EvoStar 2025 – The Leading European Event on Bio‑Inspired AI

Trieste, Italy (and Hybrid), 23-25 April 2025

Invited Speakers

GECCO 2025

The Genetic and Evolutionary Computation Conference (GECCO 2025) presents the latest high-quality results in genetic and evolutionary computation since 1999.

Málaga, Spain (and Hybrid), July 14 – 18, 2025

Keynotes

Call for Papers

GECCO 2025 Calls for Hot-off-the-Press, Late Breaking Abstracts and Competition Entries

2025 Genetic and Evolutionary Computation Conference (GECCO 2025)
14 – 18 July 2025, Málaga, Spain

Calls for Submissions

The submission deadline is 12 April 2025. Looking forward to receiving your submissions.

18th ACM/SIGEVO Conference on Foundations of Genetic Algorithms
FOGA XVIII

Aug 27 – 29, 2025, Leiden, The Netherlands

FOGA 2025 is a conference organized by ACM/SIGEVO and hosted by the Leiden Institute of Computer Science (LIACS) in Leiden, The Netherlands. The conference will this year take place on three days (Wed – Fri).

The FOGA series aims at advancing our understanding of the working principles behind evolutionary algorithms and related randomized search heuristics, such as local search algorithms, differential evolution, ant colony optimization, particle swarm optimization, artificial immune systems, simulated annealing, and other Monte Carlo methods for search and optimization. Connections to related areas, such as Bayesian optimization and direct search, are of interest as well. FOGA is the premier event to discuss advances on the theoretical foundations of these algorithms, tools needed to analyze them, and different aspects of comparing algorithms’ performance

Keynote Speakers

Important Dates

  • Submission: May 2, 2025
  • Notification of decision to authors: June 27, 2025
  • Camera ready: July 10, 2025
  • Early registration: July 10, 2025
  • Conference: Aug 27 – 29, 2025

17th International Conference on Evolutionary Computation Theory and Applications (ECTA)

ECTA 2025 is part of IJCCI, the 17th International Joint Conference on Computational Intelligence. Registration to ECTA allows free access to all other IJCCI conferences.

Important Dates

Regular Papers

  • Paper Submission: May 19, 2025
  • Authors Notification: July 17, 2025
  • Camera Ready and Registration: July 31, 2025

Position Papers

  • Paper Submission: June 26, 2025
  • Authors Notification: July 31, 2025
  • Camera Ready and Registration: September 5, 2025

22nd Annual (2025) “Humies” Awards

For Human-Competitive Results – Produced by Genetic and Evolutionary Computation

To be held as part of the Genetic and Evolutionary Computation Conference (GECCO)
July 14-18, 2025 (Monday – Friday), Málaga, Spain (Hybrid)

Detailed Call for Entries

Entries are hereby solicited for awards totaling $10,000 for human-competitive results that have been produced by any form of genetic and evolutionary computation (including, but not limited to genetic algorithms, genetic programming, evolution strategies, evolutionary programming, learning classifier systems, grammatical evolution, gene expression programming, differential evolution, genetic improvement, etc.) and that have been published in the open, reviewed literature between the deadline for the previous competition and the deadline for the current competition.

Important Dates

  • Friday, May 30, 2025: Deadline for entries (consisting of one TEXT file, PDF files for one or more papers, and possible “in press” documentation). Please send entries to goodman at msu dot edu
  • Friday, June 13, 2025: Finalists will be notified by e-mail
  • Friday, June 27, 2025: Finalists must submit a 10-minute video or, if presenting in person, their slides, to goodman at msu dot edu.
  • July 14-18, 2025 (Monday – Friday): GECCO conference (the schedule for the Humies session is not yet final, so please check the GECCO program as it is updated for the time of the Humies session). GECCO will be in hybrid mode, so the finalists may present their entry in person or on video.
  • Friday, July 18, 2025: Announcement of awards at the plenary session of the GECCO conference.

About this Newsletter

SIGEVOlution is the newsletter of SIGEVO, the ACM Special Interest Group on Genetic and Evolutionary Computation. To join SIGEVO, please follow this link: [WWW].
We solicit contributions in the following categories:

Art: Are you working with Evolutionary Art? We are always looking for nice evolutionary art for the cover page of the newsletter.

Short surveys and position papers. We invite short surveys and position papers in EC and EC-related areas. We are also interested in applications of EC technologies that have solved interesting and important problems.

Software. Are you a developer of a piece of EC software, and wish to tell us about it? Then send us a short summary or a short tutorial of your software.

Lost Gems. Did you read an interesting EC paper that, in your opinion, did not receive enough attention or should be rediscovered? Then send us a page about it.

Dissertations. We invite short summaries, around a page, of theses in EC-related areas that have been recently discussed and are available online.

Meetings Reports. Did you participate in an interesting EC-related event? Would you be willing to tell us about it? Then send us a summary of the event.

Forthcoming Events. If you have an EC event you wish to announce, this is the place.

News and Announcements. Is there anything you wish to announce, such as an employment vacancy? This is the place.

Letters. If you want to ask or say something to SIGEVO members, please write us a letter!

Suggestions. If you have a suggestion about how to improve the newsletter, please send us an email.

Contributions will be reviewed by members of the newsletter board. We accept contributions in plain text, MS Word, or Latex, but do not forget to send your sources and images.

Enquiries about submissions and contributions can be emailed to gabriela.ochoa@stir.ac.uk
All the issues of SIGEVOlution are also available online at: evolution.sigevo.org

Notice to contributing authors to SIG newsletters
As a contributing author, you retain the copyright to your article. ACM will refer all requests for republication directly to you.

By submitting your article for distribution in any newsletter of the ACM Special Interest Groups listed below, you hereby grant to ACM the following non-exclusive, perpetual, worldwide rights:

  • to publish your work online or in print on the condition of acceptance by the editor
  • to include the article in the ACM Digital Library and in any Digital Library-related services
  • to allow users to make a personal copy of the article for noncommercial, educational, or research purposes
  • to upload your video and other supplemental material to the ACM Digital Library, the ACM YouTube channel, and the SIG newsletter site

If third-party materials were used in your published work, supplemental material, or video, make sure that you have the necessary permissions to use those third-party materials in your work

Editor: Gabriela Ochoa

Sub-editor: James McDermott

Associate Editors: Emma Hart, Bill Langdon, Una-May O’Reilly, Nadarajen Veerapen, and Darrell Whitley