In global optimization techniques, it is often the case that the data structures used to represent a candidate solution can be directly decoded and interpreted; hence, the best fitting individual immediately provides an explicit and human-readable description of the optimal solution. Research mainly focused on the creation of novel strategies, aimed at balancing the exploration and exploitation capabilities of optimization algorithms.
Although this well-established research line is very prolific—paving the way to the design of efficient algorithms even for large-scale problems—there is another promising direction that could be investigated, consisting in the modification of the search space, that is, space transformation able to dilate, shrink, stretch, collapse, smooth, or remap the fitness landscape, leading to alternative or simplified formulations of the original optimization problem.
In discrete domains, a similar approach can be performed by embedding implicit or explicit assumptions into the structure of candidate solutions, so that genetic operators can explore the feasible search space in a “smarter” way, reducing the overall computational effort. Example of this technique include generative representations, techniques based on grammars, single parent techniques where example genes are incorporated, and other structured representations that starkly limit the portion of the search domain examined.
This special session aims at gathering the researchers who are investigating new directions and ideas in the field of candidate solutions representation and its dual notion fitness landscape manipulation.
This Special Session welcomes any paper considering all kinds of non-conventional candidate solutions representation, including the dual perspective of fitness landscape manipulation. Examples include but are not limited to:
- Non-conventional representations of candidate solutions
- Dilation functions and other functions that reshape the fitness landscape
- Alternative semantics for candidate solutions
- Fitness landscape modification, simplification, and restriction
- Novel closed variation/evolutionary operators
- Implicit/relative representations
- Generative or developmental representations
- Self-adaptive representations
- Parameterized manifolds of representations
- State-conditioned representations
- Procedural representations
- Generative automata
- Surrogate models
- Highly interpretable representations
- Wendy Ashlock (Ashlock and McGuinness Consulting, Canada)
- Laurens Bliek (Eindhoven University of Technology, The Netherlands)
- Joseph A. Brown (Innopolis University, Russia)
- Paolo Cazzaniga (University of Bergamo, Italy)
- Caro Fuchs (Eindhoven University of Technology, The Netherlands)
- Sheridan Houghten (Brock University, ON, Canada)
- James Hughes (St. Francis Xavier University, NS, Canada)
- Uzay Kaymak (Eindhoven University of Technology, The Netherlands)
- Luca Manzoni (University of Trieste, Italy)
- Giancarlo Mauri (University of Milano-Bicocca, Italy)
- Eric Medvet (University of Trieste, Italy)
- Daniele Papetti (University of Milano-Bicocca, Italy)
- Simone G. Riva (Cambridge University, UK)
- Gonzalo Ruz (Universidad Adolfo Ibanez, Chile)
- Sara Silva (Universidade de Lisboa, Portugal)
- Simone Spolaor (University of Milano-Bicocca, Italy)
- Andrea Tangherloni (University of Bergamo, Italy)
- Submission deadline: 31 January 2021
- Notification: 22 March 2021
- Final paper submission: 7 April 2021
Special session papers should be uploaded online through the paper submission website of IEEE CEC 2021.
Please select the corresponding special session name (“Special Session on candidate solutions representation and fitness landscape manipulation”) as the “main research topic” in submission.
Papers in this session must both be appropriate to the session and meet the criteria for papers for IEEE CEC 2021.
- Daniel Ashlock
Professor, Department of Mathematics and Statistics, University of Guelph, ON, Canada
- Daniela Besozzi
Associate Professor, Department of Informatics, Systems and Communication, University of Milano-Bicocca, Italy
- Marco S. Nobile
Assistant Professor, Department of Industrial Engineering & Innovation Sciences, Eindhoven University of Technology, The Netherlands