The workshop WEPO 2021 is intended to foster the discussion about Evolutionary Computation (EC) and Population-based Optimization (PO). Nature-inspired algorithms are largely used for solving optimization problems in a large number of fields due to their simplicity and effectiveness. The underlying principles behind these algorithms are simple enough to allow a great adaptability to various problems and domains and, while maintaining excellent effectiveness, they offer the possibility of obtaining explainable solutions. In a scenario where AI is increasingly predominant, but often with black box solutions, the explainability of EC and PO solutions may be an answer to the growing demand for understandable AI. The goal of this workshop is to explore and discuss the latest trends, promising results and hot topics in the fields of EC and PO, offering a discussion forum where new research collaborations can be established.
The workshop is directed towards researchers, practitioners, and students that work on or are interested in evolutionary computation (EC) and population-based optimization (PO) methods. Particular emphasis is given to recent advances in the use of EC and PO in the area of explainable AI.
Download the Call for Papers
The topics of the workshops include all the evolutionary computation methods and population-based optimization techniques, including, but not limited to:
Both theoretical works, novel techniques, and application to real-world problems are on-topic for the workshop. A particular focus of this workshop is the relation between these methods and the larger area of AI, in particular how they can contribute to the development of an explainable AI and how they can be hybridized with other machine learning methods.
WEPO 2021 welcomes three different kinds of submissions:
All submissions must be formatted according to Springer’s LNCS style, and must be submitted using the EasyChair submission system by selecting the WEPO track.
Paper submission deadline (FINAL): Oct 8, 2021 Oct 15, 2021
Notification to authors: Nov 10, 2021 Nov 12, 2021
Camera-ready submission: Nov 22, 2021
Workshop date: Nov 30, 2021
The workshop will be held online on Nov 30, 2021, before the main conference. For further information about participation and registration, please refer to the AIxIA 2021 website.
All accepted research and exploratory papers will be available in a local proceedings volume or, subject to confirmation, to CEUR-WS proceedings. All accepted software demos will be available in the local proceedings.
Authors of a selection of accepted papers will be invited to submit an extended and revised version for publication in an international journal (to be confirmed).
The program starts at 9:00am, CEST time (UTC + 2). Please check the AIxIA schedule on UNDERLINE for the Zoom link to attend the workshop.
TIME CEST (UTC+2) |
SESSION/TITLE |
---|---|
9:00 - 9:05 | Welcome note (Andrea De Lorenzo) |
Session 1: Invited talk (Chair: Luca Mariot) |
|
9:05 - 9:50 | Invited talk 1: On the Applications of Evolutionary Algorithms to the Cybersecurity Domain Stjepan Picek, Radboud University, The Netherlands [slides] |
9:50 - 10:00 | Break |
Session 2: Contributed talks (Chairs: Luca Mariot and Luca Manzoni) |
|
10:00 - 10:30 | Integrating Grammatical Evolution with Neural Fitness Functions Claudio Ferretti, University of Milano-Bicocca, Italy [paper | slides] |
10:30 - 11:00 | Application of a genetic algorithm for solving the Dial-a-Ride problem Marko Djurasevic, University of Zagreb, Croatia [paper | slides] |
11:00 - 11:30 | Deriving Smaller Orthogonal Arrays from Bigger Ones with Genetic Algorithms Luca Mariot, Delft University of Technology, The Netherlands [paper | slides] |
11:30 - 12:00 | Swarm Intelligence Algorithms for Convolutional Neural Networks Eva Tuba, Singidunum University, Serbia [paper | slides] |
12:00 - 12:10 | Break |
Session 3: Invited talk (Chair: Luca Manzoni) |
|
12:10 - 12:55 | Invited talk 2: Building the Building Blocks - Evolving Hyper-Features for Model Robustness and Readability Sara Silva, University of Lisbon, Portugal |
12:55 - 13:00 | Closing remarks (Andrea De Lorenzo) |
Evolutionary algorithms (EAs) are successfully applied in many application domains. One of those domains is cybersecurity. This talk will cover several cybersecurity applications where evolutionary algorithms showed competitive performance. First, we will discuss evolutionary algorithms and their applications for cryptography (evolution of cryptographic primitives and side-channel analysis). Afterward, we will cover EAs applications for intrusion detection, modeling attacks of Physically Unclonable Functions, fuzzing, and adversarial machine learning. Finally, we will conclude this talk with a brief overview of common challenges and possible future research directions.
Stjepan Picek is an associate professor at Radboud University, Nijmegen, The Netherlands. His research interests are security/cryptography, machine learning, and evolutionary computation. Before the associate professor position, Stjepan was assistant professor at TU Delft, The Netherlands, a postdoctoral researcher at MIT, USA, and a postdoctoral researcher at KU Leuven, Belgium. Stjepan finished his PhD in 2015 with a topic on cryptology and evolutionary computation techniques. Stjepan also has several years of experience working in industry and government. Up to now, he gave more than 20 invited talks at conferences and summer schools and published more than 100 refereed papers in both evolutionary computation and cryptography journals and conferences. Stjepan is a member of the organization committee for International Summer School in Cryptography and a general co-chair for Eurocrypt 2021.
Among its many competences, Genetic Programming (GP) can also be regarded as a feature discovery method. M3GP is a GP variant originally developed for performing multiclass classification, that recently proved to be also a powerful method for evolving hyper-features from the original data, for both classification and regression, and for a variety of machine learning methods. This talk will address how the evolved hyper-features can improve the robustness and readability of data models, with examples from remote sensing applications.
Sara Silva is a Principal Investigator of the LASIGE research center at the Faculty of Sciences of the University of Lisbon, Portugal. Her research interests are mainly in machine learning with a strong emphasis in Genetic Programming, where she has contributed with several new methods and applied them in projects related to such different domains as remote sensing, biomedicine, systems biology, maritime security, plant phenotyping, ecotoxicology and radiomics, among others. She is the author of around 100 peer-reviewed publications, and has received more than 10 nominations and awards for best paper and best researcher. In 2018 she received the EvoStar Award for Outstanding Contribution to Evolutionary Computation in Europe. She is the creator and developer of GPLAB - A Genetic Programming Toolbox for MATLAB.
Mauro Castelli, Universidade Nova de Lisboa, Portugal
Eric Medvet, University of Trieste, Italy
Laura Nenzi, University of Trieste, Italy
Marco S. Nobile, Eindhoven University of Technology, the Netherlands
Sara Silva, Universidade de Lisboa, Portugal
Leonardo Trujillo, Instituto Tecnológico de Tijuana, Mexico
Eva Tuba, Singidunum University, Serbia
Marco Virgolin, Centrum Wiskunde & Informatica, the Netherlands