Evolutionary Computation for the Design of General Purpose Artificial Intelligent Systems (GPAIS)

Within the field of GPAIS, several research areas contribute to the design of AI models with greater versatility, facilitating the learning of new tasks. These areas include, but are not limited to Multi-task models, which handle multiple tasks simultaneously; few-shot learning models, which require fewer training data points to understand and perform new tasks; auto machine learning (AutoML) and neural architecture search (NAS) systems, which can automatically adapt the AI model to a new problem; and models adaptation for new problems capable of taking advantage of prior knowledge. Collectively, these research areas are making progress in the development of GPAIS models. The flexibility and ease of adaptation of EC make them a perfect match to cope with the stringent properties sought for GPAIS, including the multimodality of the tasks being solved, their variability over time or the large dimensionality of the design and construction of GPAIS.  Actually, there are many research areas in EC that can be useful in either designing or enriching these AI models. However, the research in this line is often developed in parallel, without communication channels between them, which would allow a global vision of how EC can improve the design of these increasingly generic models.

In this Special Session, we encourage researchers to submit works that exploit EC to improve the creation or improvement of GPAIS.  The hybridization of this family of solvers with GPAIS can produce a bright future in the field of AI. More specifically, we encourage interested researchers to submit their original work on this topic. The list of potential themes includes, but is not limited to:

•  Evolutionary neural architecture search, in particular to enforce robustness and/or adaptation to new problems.

•  EC for Auto Machine Learning and/or meta-learning

•  EC for Algorithm Construction and Design

•  EC for multi-tasks

•  Evolutionary Reinforcement Learning

•  EC to learn with very few data: few-shot learning and zero-shot learning

•  Cooperative and collective learning in GPAIS

•  EC for Continual Learning

•  Improvement of the learning process: Evolutionary data generation, evolutionary data pre-processing, .…

•  Any other contributions that exploit EC to either design or enhance GPAIS

Paper Submission Deadline: January 15, 2025

Please submit your paper directly through the IEEE CEC2025 OpenReview submission system, selecting this special session as the main research topic.

For paper guidelines, please visit https://www.cec2025.org/index/page.html?id=1298

For submissions, please select the single topic "Special Session: GPAIT2: General Purpose Artificial Intelligence Technologies and Trustworthiness" from the "Special Session Papers" on https://cmt3.research.microsoft.com/IJCNN2025/.

Submitted papers will be peer-reviewed with the same criteria as other IEEE CEC 2025 tracks.

The papers accepted for the special session will be included in the IEEE CEC 2025 proceedings and will be published by the IEEE Xplore Digital Library.

Prof. Dr. Isaac Triguero

Dept. of Computer Science and Artificial Intelligence.

DaSCI, Andalusian Research Institute in Data Science and Computational Intelligence

University of Granada, Spain

triguero@decsai.ugr.es

Google Scholar

Prof. Dr. Daniel Molina

Dept. of Computer Science and Artificial Intelligence.

DaSCI, Andalusian Research Institute in Data Science and Computational Intelligence

University of Granada, Spain

dmolina@decsai.ugr.es

Prof. Dr. Bing Xue

Victoria University of Wellington 

School of Engineering and Computer Science, 6012 Wellington, New Zealand

bing.xue@ecs.vuw.ac.nz

Prof.  Dr. Ferrante Neri

School of Computer Science and Electronic Engineering

University of Surrey, Guildford, Surrey, United Kingdom

f.neri@surrey.ac.uk