SIGEVOlution
newsletter of the ACM Special Interest Group on Genetic and Evolutionary Computation
“Group photo” after the GECCO 2021 closing ceremony showing a screenshot of the participants in the virtual room created with the Gather web-conferencing software.
Editorial
Welcome to the Autumn 2021 issue of the SIGEvolution newsletter! We start with a summary of the Genetic and Evolutionary Computation Conference, GECCO 2021, which was held virtually in July with the local organising team based in Lille, France. We continue with an overview of the 18th Annual Human-Competitive Results Awards (Humies), a competition that awards $10,000 in cash prizes for computational results that are deemed to be competitive with results produced by human beings, but are automatically generated with evolutionary algorithms. Our third contribution features the results of the 2021 SIGEVO best dissertation award, which was created in 2019 to recognize excellent thesis research by doctoral candidates in the field of evolutionary computation. We conclude by announcing recent awards, events and forthcoming calls for papers.
Please get in touch if you would like to contribute an article for a future issue or have suggestions for the newsletter.
Gabriela Ochoa, Editor.
GECCO 2021: Summary and Statistics
Francisco Chicano, University of Malaga, Spain
The 2021 Genetic and Evolutionary Computation Conference (GECCO 2021) was planned to be held in hybrid mode in Lille, France, on July 10th-14th, 2021. After a careful analysis of the evolution of the pandemic and the results of a poll to frequent attendees on the preferred form of attendance, a decision was made in mid-December to run GECCO’21 in online/virtual mode only. This report provides statistics about submissions and authorship, and some comments about the evolution and growth of GECCO.
GECCO 2021 would not have been possible without the efforts of a team masterfully guided by Krzysztof Krawiec (General Chair). Bilel Derbel (Local Chair), Nadarajen Veerapen (Electronic Media Chair) and Arnaud Liefooghe (Virtualization Chair) were in charge of setting up the main virtual platforms (Gather, Whova and Zoom) that made the event possible. Alberto Tonda (Proceedings Chair), Aniko Eckart (Publicity Chair), Roxane Rose (Executive Events), Brenda Ramírez (Executive Events) and all the organizers, track chairs and program committee members completed the team.
The GECCO paper submission process has two stages. In the first stage the authors have to send an abstract of the paper and, one week later, in the second stage the authors have to upload the full version of the paper. GECCO 2021 received a total of 431 abstracts. 362 of them were submitted as full papers and 134 of those submissions were accepted as full papers, giving an acceptance rate of 37%, which is typical of high-quality conferences. From the submissions that were not accepted as full papers, 166 were accepted as posters. There were also 28 poster-only submissions and 23 of them were accepted for the poster session.
This year there were two important changes in the track structure. First, the topics of the Digital Entertainment Technologies and Arts (DETA) track were included in the Real World Applications track. Second, a new track was created: the Neuroevolution track. GECCO 2021 was organized into these 13 main tracks: Ant Colony Optimization and Swarm Intelligence (ACO-SI), Complex Systems (Artificial Life/Artificial Immune Systems/ Generative and Developmental Systems/Evolutionary Robotics/Evolvable Hardware) (CS), Evolutionary Combinatorial Optimization and Metaheuristics (ECOM), Evolutionary Machine Learning (EML), Evolutionary Multiobjective Optimization (EMO), Evolutionary Numerical Optimization (ENUM), Genetic Algorithms (GA), General Evolutionary Computation and Hybrids (GECH), Genetic Programming (GP), Neuroevolution (NE), Real World Applications (RWA), Search-Based Software Engineering (SBSE) and Theory.
The number of submissions (yellow), the number of accepted full papers (blue), the acceptance rate of each track (green) and the acceptance rate of the conference (red) are shown in Fig 1. The tracks are sorted by the number of full paper submissions they received (decreasing from left to right). The RWA track received the largest number of submissions. The acceptance rate is different in each track, with slight deviations around the average in most of the cases. In the case of the tracks with fewer submissions, Theory and SBSE, these deviations are high because one paper more or less, here, gives a large change in the acceptance rate. The largest acceptance rate was 64% in the Theory track, while the lowest acceptance rate was 27% in the SBSE track.
Figure 1. GECCO2021 full paper submission and acceptance rate per track.
Figure 2. List of countries of 80% of the authors.
The submissions received at GECCO 2021 were authored by 1025 researchers from 52 countries with 80% of authors affiliated to just 17 countries (see Fig. 2). One remarkable fact is that China is in the first position in the ranking of authors’ countries with 123 authors. In the second position we find the USA (119), then Germany (86), and the UK (84). France, the hosting country, is the fifth in the ranking with 65 authors.
Figure 3. The number of authors per country (grouped by continent).
Fig. 3 contains a visualization of the authors per country and continent. The area of the circles is proportional to the number of authors per country. Circles are colored and grouped by continent. There are few submissions from South America and Africa, which suggests opportunities for growing the GECCO community
In addition to the submissions to the main conference, GECCO 2021 received 121 papers for 22 specialized workshops, 9 submissions in the Student Workshop, 11 Late-Breaking Abstracts and 24 Hot Off the Press submissions. GECCO 2021 also provided 38 tutorials (introductory, advanced and specialized) from among 45 proposals.
This edition of GECCO was very fortunate to have three fantastic keynote speakers: Joshua Tenenbaum (MIT, USA), Marc Mézard (École Normale Supérieure, France) and Melanie Mitchell (Santa Fe Institute, USA).
Historical evolution of GECCO
Figure 4. Full-paper submissions and acceptance rate for past GECCO editions.
Comparing the number of submissions and acceptance rate of GECCO 2021 to its previous editions (Fig. 4), it can be inferred that the number of submissions strongly depends on the location of the conference. The acceptance rate of GECCO 2021 is below 40%, the same as in the last 10 years, in the range of 33% to 38%. Regarding the number of submissions, it is slightly lower than in previous editions, probably due to the effects of the pandemic. We should not forget that the GECCO deadline was during one of the strongest waves of COVID-19 in Europe, the hosting continent.
Figure 5. The number of accepted full papers per track and year. The width of each line gives the relative size of the track, while the vertical position gives the ranking in terms of size (larger tracks are shown near the top).
The evolution of the main GECCO tracks is shown in Fig. 5, where the (vertical) width of each line is relative to the size of the track (number of accepted full papers) and the vertical position of the line gives the relative order (larger to smaller, from top to bottom).
Finally, the growth of GECCO is also clearly illustrated by the number of registrations, which achieved a historical maximum of 946 in GECCO 2021. The record of registrations for an onsite edition is 691, reached in GECCO 2018 (Kyoto, Japan). Offering reduced registration fees for non-presenters contributed to having attendees that do not normally go to GECCO, and helped to increase the visibility of the conference in more countries. These attendees can still watch some of the pre-recorded videos of the talks, keynotes, and plenary sessions, and participate in the discussion chats for at least the next 6 months.
Impacted by COVID-19, GECCO 2021 had to use new tools to provide networking and socialization among the attendees. Gather and Whova played that role. While we all miss sitting at a real table with some drinks to discuss research ideas, Gather tried to fill that gap by providing virtual tables, rooms, whiteboards, and a “cam-to-cam” interface. We hope attendees enjoyed these tools and felt a bit closer during the conference.
The GECCO Conference, held virtually this year in Lille, France, July 10-14, was the home for the 18th Annual Human-Competitive Results Awards. The Humies competition annually awards $10,000 in cash prizes for computational results that are deemed to be competitive with results produced by human beings, but are generated automatically by computer. This year, there were 20 entries submitted, and the judges selected six of those to be finalists, based on the papers and entry forms they submitted. Because GECCO was virtual, the six Humies finalists did not present live talks, but instead submitted 10-minute videos that were shown during the Humies session and viewed by the judges in making their selections of the prize winners. All 20 entries and the six finalists’ videos are all available to the general public at the website for the Humies, www.human-competitive.org , so anyone can watch them there.
The competition, sponsored by John Koza (who is widely regarded as the “Father of Genetic Programming”) solicits papers published within the last year that describe work fulfilling one or more of eight criteria, including such features as winning a regulated competition against humans or other programs, producing results that are publishable in their own right (not because they were created by a computer program) patentability, and other criteria described in full on the Humies website. The Gold Award includes US$5,000; the Silver Award, $3,000, and the Bronze, $2,000. This year’s judges for the competition were Wolfgang Banzhaf, Stephanie Forrest, Erik Goodman, Una-May O’Reilly, Lee Spector and Darrell Whitley. Publicity for the Humies was done with excellent results by Bill Langdon.
As usual, after considering all of the entries, the judges wanted to give awards to many of them, but the final result was the following three awards:
Gold Award
The Gold and a $5,000 prize was awarded to a team including Esteban Real, Chen Liang, David R. So, and Quoc V. Le, all from Google Research in Mountain View, CA. Their winning paper was entitled “AutoML-Zero: Evolving Machine Learning Algorithms From Scratch,” and was published in the Proceedings of the 2020 ICML Conference. Rather than trying to evolve the architecture of a deep neural network via Network Architecture Search (NAS), as has been done very successfully in recent years, they set out to see if evolutionary search could discover the basic algorithms of machine learning, starting with operators like addition, multiplication, etc.—none that a high school math class wouldn’t cover—and with scalars, vectors, and matrices as data structures. Their evolution begins with three EMPTY functions called in the supervised learning framework (i.e., that is how the evolution proceeds and correct classifications are rewarded). The functions are Setup(), Predict(), and Learn(), and they evolve to contain sets of instructions made up of those primitives mentioned above. They don’t provide things like derivatives—if a derivative is wanted, the algorithm for computing it must be evolved. They showed that using simple classification tasks in a supervised learning framework, they could evolve two-layer neural networks trained by backpropagation, and many enhancements such as bilinear interactions, normalized gradients, and weight averaging. After the classifier is evolved (which includes any training steps), they look at the resulting code and manually try to determine what the algorithm is doing at each stage of the evolution, and have discovered that many of the basic ML and NN algorithms have been evolved. Their results and their code are all available on GitHub. Here is the video presentation.
A slide from the video presentation by Steven Real of the entry receiving the Gold Award at the Humies 2021.
Silver Award
The Silver Award and $3,000 went to a large team, Marco Virgolin, Ziyuan Wang, Tanja Alderliesten, Peter A.N. Bosman, Brian Balgobind, Irma W.E.M. van Dijk, Jan Wirsma, Petra Kroon, Gert Janssens, Marcel van Herk, David Hodgson, Lorna Zadravec Zaletel, Coen Rasch, and Arjan Bel, most of whom are from the Netherlands, but with one each from Slovenia, UK, and Canada. Their lead paper was entitled “Machine learning for the prediction of pseudorealistic pediatric abdominal phantoms for radiation dose reconstruction.” Their work is aimed at imputing the radiation dose that was received by various non-target organs during radiation treatment of children many years ago, in order that the side effects they are experiencing much later can be related to the doses of radiation received by the affected organs. For this, they use available 2-D imaging to parameterize 3-D “phantoms” representing those organs, by training a variety of machine learning algorithms to create those phantoms, then determine the dosages they would have received under the treatment regime used. One of the ML methods significantly outperforms both the other algorithms and the human-created phantoms, promising to enable more accurate future determination of adverse effects as a function of radiation dosage. Here is the video presentation.
A Slide from the presentation by Marco Virgolin of the entry receiving the Silver Award at the Humies 2021.
Bronze Award
The Bronze Award and $2,000 went to Daniel Blasco, Jaime Font, Mar Zamorano, and Carlos Cetina, all of Universidad San Jorge, Zaragoza, Spain. Their entry was entitled “An Evolutionary Approach for Generating Software Models: The Case of Kromaia in Game Software Engineering.” They used an Evolutionary Model Generation (EmoGen) approach to generate software models comparable in quality to the models created by human developers, in a tiny fraction of the time—humans take ten months to create a “boss” for the Kromaia game that is appropriate in interest and difficulty for a level in the game, whereas their program generated equivalently challenging bosses in five hours. These bosses each have more than 1000 model elements, so the problem is challenging. Here is the video presentation.
A slide from the presentation by Jaime Font of entry receiving the Bronze Award at the Humies 2021.
More Information
See any of the finalists’ videos and papers and entry forms at www.human-competitive.org. And start preparing your own best human-competitive work for the 2022 Humies competition at GECCO next July! Any questions? Contact me, Erik Goodman, goodman@msu.edu.
ACM SIGEVO Best Dissertation Award 2021
Manuel López-Ibáñez, University of Málaga, Spain
The SIGEVO Best Dissertation Award was created in 2019 to recognize excellent thesis research by doctoral candidates in the field of evolutionary computing. Doctoral dissertation awards are given by other Special Interest Groups of ACM, such as SIGCOMM, SIGKDD, SIGARCH, and others. The SIGEVO Best Dissertation Award will be given annually to a maximum of 1 winner and a maximum of 2 honorable mentions. The award presentation will take place at the awards ceremony of GECCO. The award carries a monetary value of $1,000 contributed by SIGEVO to be awarded to the winner. The award winner and honorable mentions will each receive a plaque.
Dissertations are reviewed by a selection committee for the technical depth and significance of the research contribution, the potential impact on the field of evolutionary computing, and the quality of presentation. This year, the members of the selection committee were:
Thomas Bäck
University of Leiden, The Netherlands
Francisco Chicano
University of Málaga, Spain
Kenneth De Jong
George Mason University, USA
Jonathan Fieldsend
University of Exeter, UK
Manuel López-Ibáñez
University of Málaga, Spain
Tea Tušar
Jožef Stefan Institute, Slovenia
The committee received and reviewed 6 nominations this year, in topics as diverse as multi-objective optimization, evolutionary design, theory of evolutionary algorithms, evolutionary clustering, computational biology, and genetic programming.
The winner of the 2021 SIGEVO Best Dissertation Award is “Design and Application of Gene-pool Optimal Mixing Evolutionary Algorithms for Genetic Programming” by Marco Virgolin from Delft University of Technology (The Netherlands) with the following citation from the selection committee:
“This dissertation introduces significant innovations in model-based genetic programming (GP) enabling the efficient discovery of compact and accurate solutions, which is key for supporting explainable Artificial Intelligence. In collaboration with medical researchers, the proposed innovations are applied successfully to improve historical 3D-dose reconstruction in radiation oncology which is used to design more effective treatments for cancer patients. This work is a stellar example of the positive real-world impact of Evolutionary Computation.”
In addition to the winner, the committee decided to award two honorary mentions due to their high quality.
The first honorary mention is given to “Methods for Tight Analysis of Population-based Evolutionary Algorithms” by Denis Antipov from the Institut Polytechnique de Paris (France) with the following citation:
“This dissertation significantly advances the state of the art in our mathematical understanding of how the parameters of an evolutionary algorithm influence its performance. Four novel analytical tools are proposed that have the potential of leading to new developments in the theory of evolutionary computation, beyond the results already obtained here. Among other results, the analysis in this work shows that heavy-tailed random parameter choices can give excellent performance over static parameter choices even for more complicated algorithms with several parameters. This result suggests elegant algorithmic strategies to address the notoriously difficult problem of online parameter control.”
The second honorary mention was given to “Accelerating Evolutionary Design Exploration with Predictive and Generative Models” by Adam Gaier from the Université de Lorraine (France) with the following citation:
“The dissertation introduces new algorithms and a novel data-driven encoding approach that vastly improve the efficiency and real-world applicability of quality-diversity (QD) algorithms. The proposals represent foundational work on the integration of quality-diversity (QD) and machine learning (ML). These proposals are evaluated in real-world examples including aerodynamics design. The results are an important step in engineering and industrial design exploration.”
The next edition of the award will welcome dissertations defended between January and December of 2021. The deadline for submitting nominations is February 1st, 2022. More information is available at: https://sig.sigevo.org/index.html/SIGEVO+Dissertation+Award
We encourage you to disseminate this information among your colleagues and students! We are looking forward to your nominations!
Darrell Whitley Awarded as the 2022 Evolutionary Computation Pioneer
Prof. Darrell Whitley has received the 2022 IEEE Pioneer Award on Evolutionary Computation from the IEEE Computational Intelligence Society “For early contributions to the practical implementation and theoretical understanding of evolutionary algorithms.” By definition IEEE Pioneer Awards are for contributions made more than 15 years ago.
We asked Darrell: What do you see as some of your early contributions?
Here is his answer:
“One of my early contributions was the development of the Steady State Genetic Algorithm, which was supported by the release of the GENITOR programming environment for genetic algorithms. GENITOR first appeared in 1988. GENITOR was widely downloaded and used in the 1990s. The name steady-state genetic algorithm came after GENITOR, during a lunch between ME, Dave Davis, and Gil Sysweda in 1990 at the first FOGA workshop. Another important aspect of GENITOR was the use of ranked-based selection. Virtually all genetic algorithms used fitness proportional selection at the time. "
"I also did some of the earliest work applying genetic algorithms to real-world resource scheduling problems. I have continued to work on real-world scheduling applications over my entire career.
From a theoretical point of view, I developed one of the first infinite population models of the standard genetic algorithm and presented a paper on the model at the FLAIRS conference in the fall of1990. Michael Vose and Gunar Liepins were independently developing their own model about the same time and came out with a paper in 1991.
Working with my students, we also showed in 1998 that it was possible to exactly compute hyperplane averages in closed form for k-bounded problems such as MAX-3SAT.
This proved that one cannot rely on hyperplane averages to infer globally optimal solutions. One still sees papers that make exaggerated claims about the optimal sampling properties of genetic algorithms that are not supported by the theory. "
Joint Lectures on Evolutionary Algorithms (JoLEA)
The Centrum Wiskunde & Informatica (CWI)’s Peter A.N. Bosman, Arkadiy Dushatskiy and Marco Virgolin, together with Leiden University's Thomas Bäck, Anna Kononova, and Jacob P. de Nobel, Utrecht University's Dirk Thierens, and Vrije University of Amsterdam's Gusz Eiben, proudly present the Joint Lectures on Evolutionary Algorithms (JoLEA).
From left to right, top to bottom: Thomas Bäck (Leiden University), Peter A.N. Bosman (CWI), Gusz Eiben (Vrije Universiteit Amsterdam), Dirk Thierens (Utrecht University), Arkadiy Dushatskiy (CWI), Anna Kononova (Leiden University), Jacob P. de Nobel (Leiden University), Marco Virgolin (CWI).
JoLEA is an initiative that was born in 2021 in the Netherlands, aimed at disseminating high-quality research on evolutionary algorithms. JoLEA events take place every two months. Speakers alternate between world-renowned senior researchers, who provide a lecture on their established research line, and junior researchers (PhD students, postdocs, etc.), who present and discuss their latest and work-in-progress research results.
Anton Bouter (left), Roy de Winter (right), and images representing the applications tackled by their EAs (middle)
The first lecture took place on 15-09-2021, featuring scientific software developer and PhD candidate Anton Bouter from CWI with “Optimal Mixing Evolutionary Algorithms for Large-Scale Real-Valued Optimization”, and PhD candidate Roy de Winter from Leiden University with “Constrained Multi-Objective Ship Design Optimization with a Limited budget of Function Evaluations”.
The next lecture is planned for November 4th, featuring Prof. Darrell Whitley on the topic of “Next Generation Genetic Algorithms”.
For more info and to stay up to date with the latest news, visit the JoLEA website You can also find JoLEA on Twitter and JoLea on LinkedIn.
Calls for Papers
EvoStar 2022
EvoStar 2022 will be held on April 20 - 22. Evostar combines four co-located conferences run every Spring. The events originated from workshops organized by EvoNet, the Network of Excellence in Evolutionary Computing, established by the Information Societies Technology Programme of the European Commission, and they represent a continuity of research collaboration stretching back over 20 years.
EvoStar is organized by SPECIES, the Society for the Promotion of Evolutionary Computation in Europe and its Surroundings. This non-profit academic society is committed to promoting evolutionary algorithmic thinking, with the inspiration of parallel algorithms derived from natural processes. It provides a forum for information and exchange.
The four conferences include:
● EuroGP: The 25th European Conference on Genetic Programming
● EvoApplications: The 24th European Conference on the Applications of Evolutionary and bio-inspired Computation
● EvoCOP: The 22nd European Conference on Evolutionary Computation in Combinatorial Optimisation
● EvoMUSART: The 11th International Conference on Evolutionary and Biologically Inspired Music, Sound, Art and Design
Important dates
● Submission deadline: November 1st, 2021
Conference: April 20th to 22nd, 2022
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 to 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: www.sigevolution.org
Notice to contributing authors to SIG newsletters
By submitting your article for distribution in the Special Interest Group publication, you hereby grant to ACM the following non-exclusive, perpetual, worldwide rights:
to publish in print on condition of acceptance by the editor
to digitize and post your article in the electronic version of this publication
to include the article in the ACM Digital Library
to allow users to copy and distribute the article for noncommercial, educational or research purposes
However, as a contributing author, you retain copyright to your article and ACM will make every effort to refer requests for commercial use directly to you.
Editor: Gabriela Ochoa
Sub-editor: James McDermott
Associate Editors: Emma Hart, Una-May O'Reilly, Nadarajen Veerapen, and Darrell Whitley