2024 IEEE Conference on Artifical Intelligence workshop on frontiers of Evolutionary Computation techniques
Workshop Introduction
Overview: Evolutionary computation is a family of algorithms modeled on natural evolution. It applies mechanisms like mutation, crossover, and selection to a population of candidate solutions, allowing them to evolve over generations. This approach excels in exploring vast search spaces and addressing complex optimisation tasks, making it a powerful tool in fields like engineering and artificial intelligence.
Traditionally, the category of evolutionary computation encompasses several sub-categories and techniques. Here is a list of some of the main categories within this field:
1. Genetic Algorithms (GA)
2. Evolution Strategies (ES)
3. Genetic Programming (GP)
4. Swarm Intelligence Algorithms
5. Differential Evolution (DE)
6. Ant Colony Optimisation (ACO)
7. Artificial Immune Systems (AIS), etc.
The dynamism and advanced research fields of evolutionary computation are vast and diverse, reflecting the ever-evolving nature of this field. Here is a list of some of the current dynamic and advanced research areas within evolutionary computation:
• Hybridisation and Integration with Other Methods
• Multi-objective Optimisation
• Large-scale Optimisation
• Constrained Optimisation
• Evolutionary Multitasking and Transfer Optimisation
• Evolutionary Deep Learning
• Evolutionary Theory
• Real-world Applications, etc.
These are just a few examples of the dynamism and advanced research fields within evolutionary computation. As the field continues to evolve, new challenges and opportunities will arise, driving further advancements in these and other areas. This workshop aims to delve into the contemporary trends of evolutionary computation techniques, explore its successes and challenges, and contemplate its future directions and broader impact within the AI community, industry, and society.
Organisers:
Zexuan Zhu (Shenzhen University, China)
Liang Feng (Chongqing University, China)
Yaqing Hou (Dalian University of Technology, China)
Ran Cheng (Southern University of Science and Technology, China)
Zhihui Zhan (South China University of Technology, China)
Overview: Neural Architecture Search (NAS) has emerged as a revolutionary approach to automating the intricate design of Artificial Neural Networks (ANNs). Over the past decade, NAS has captivated the attention and enthusiasm of both the Computational Intelligence and Machine Learning communities.
Traditionally, NAS algorithms are grouped into three main categories based on their search methodologies:
Reinforcement Learning
Gradient-based Techniques
Evolutionary Computation
Yet, the landscape of NAS is continually evolving. Modern NAS innovations often defy these clear-cut classifications. Some algorithms exhibit characteristics spanning multiple categories, while others introduce groundbreaking architectural encodings that challenge conventional search methods.
The dynamism of this field is undeniable. Contemporary NAS techniques are pushing boundaries by enabling the design of complex Deep Neural Networks within stringent time constraints. This progress is driven by:
Streamlined Search Spaces: Efficiently defined to focus on optimal solutions.
GPU Optimisation: Leveraging the power of GPUs for accelerated computations.
Approximation Techniques: Using intelligent approximating functions to enhance performance.
Notably, while some NAS approaches excel in specific domains, others produce adaptable architectures, empowering users to effortlessly transition between diverse applications.
This workshop aims to delve into the contemporary trends of NAS, explore its successes and challenges, and contemplate its future directions and broader impact within the AI community, industry, and society.
Organisers:
Zexuan Zhu (Shenzhen University, China)
Liang Feng (Chongqing University, China)
Yaqing Hou (Dalian University of Technology, China)
Ran Cheng (Southern University of Science and Technology, China)
Zhihui Zhan (South China University of Technology, China)
Overview:
In our rapidly changing world, the ability to learn and adapt is paramount. This capability enables us to navigate complex challenges, leveraging our past experiences to avoid previous pitfalls and apply knowledge in novel contexts. Remarkably, the human capacity for selecting and generalising relevant experiences to new problems stands as a testament to our cognitive sophistication.
Within the context of computational intelligence, several core learning technologies in neural and cognitive systems, fuzzy systems, probabilistic and possibilistic reasoning, have shown promise in emulating the generalisation capabilities of human learning, with many now used routinely to enhance our daily lives. Recently, in contrast to traditional machine learning approaches, Transfer Learning, which uses data from related source tasks to augment learning in a new (target) task, has attracted extensive attention and demonstrated great success in a wide range of real-world applications, including computer vision, natural language processing, speech recognition, etc.
In spite of several advances in computational intelligence, it is noted that the attempts to emulate such cognitive capabilities in problem solvers, especially those of an evolutionary nature, have received far less attention. In fact, most existing evolutionary algorithms (EAs) remain ill-equipped to exploit the potentially rich sources of knowledge that may be embedded in previous searches. With this, and the observation that any practically useful industrial system is likely to face a large number of (possibly repetitive) problems over a lifetime, it is contended that novel research advances in both the theory and application of Transfer Learning, especially for Optimisation, are primed to bring about a new wave in the real-world impact of intelligent systems.
The main goal of this workshop is to promote the research on crafting novel algorithm designs as well as theoretical analysis towards “intelligent” evolutionary computation, which possesses the transfer capabilities that evolve along with the problems solved. Further, this workshop also aims at providing a forum for academic and industrial researchers to explore future directions of research and promote evolutionary computation techniques to a wider audience in the society of computer science and engineering.
Organisers:
Zexuan Zhu (Shenzhen University, China)
Liang Feng (Chongqing University, China)
Yaqing Hou (Dalian University of Technology, China)
Ran Cheng (Southern University of Science and Technology, China)
Zhihui Zhan (South China University of Technology, China)
Keynotes Introduction
Abstract: Trustworthiness is a critical issue in artificial intelligence (AI), especially for real-world AI applications. It is impossible to apply AI in the real-world without its being trustworthy. However, the connotation and extension of trustworthiness are not entirely clear to the scientific community. There has not been a single definition that is accepted by all researchers. Nevertheless, the vast majority of researchers agree that AI trustworthiness should include at least accuracy, reliability, robustness, safety, security, privacy, fairness, transparency, controllability, maintenability, etc. This talk first reviews briefly concepts like trust, trustworthiness and trustworthy systems in engineering. Then we discuss what might be new in trustworthy AI. It is argued that AI ethics is a crucial part of trustworthy AI, which was not highlighted in traditional trustworthy systems. We present a hirarchical and multi-dimensioanl view of AI ethics, including fairness, transparency, safety, etc. Finally, we focus on the latest technical solutions to some trustworthy issues in machine learning, including fairness, explanability and safety. It is shown how multi-objective evolutionary learning can be used as an effective approach to enhancing AI trustworthiness.
Biosketch of the speaker: Xin Yao is Vice-President (Research and Innovation) and Tong Tin Sun Chair Professor of Machine Learning at the Lingnan University, Hong Kong SAR. He is an IEEE Fellow and was a Distinguished Lecturer of the IEEE Computational Intelligence Society (CIS). He served as the President (2014-15) of IEEE CIS and the Editor-in-Chief (2003-08) of IEEE Transactions on Evolutionary Computation. His major research interests include evolutionary computation, neural network ensemble learning, and real-world applications. Recently, he has been working on trustworthy AI, especially on fair machine learning and explainable AI, and AI ethics. His research work won the 2001 IEEE Donald G. Fink Prize Paper Award; 2010, 2016 and 2017 IEEE Transactions on Evolutionary Computation Outstanding Paper Awards; 2011 IEEE Transactions on Neural Networks Outstanding Paper Award; 2010 BT Gordon Radley Award for Best Author of Innovation (Finalist); and other best paper awards at conferences. He received the 2012 Royal Society Wolfson Research Merit Award, 2013 IEEE CIS Evolutionary Computation Pioneer Award and the 2020 IEEE Frank Rosenblatt Award.
Abstract:Secure and federated data-driven optimization is an emerging research area that aims to protect the data security and privacy used in optimization. This talk starts with an introduction to basic ideas of data-driven optimization and federated privacy-preserving data-driven optimization. To protect the privacy of both offline and online data, we introduce a secure federated data-driven optimization framework based on the Diffie-Hellman protocol, in which a semi-honest client is randomly chosen to solve the acquisition function and determine the next sample point, making sure that newly sampled data is also protected. To reduce the negative impact of the noise added in differential privacy, a utility function is proposed to optimize the noise level that can optimally balance privacy preservation and optimization performance. Finally, a federated multi-tasking data-driven optimization algorithm is presented that shares the hyperparameters of Gaussian processes for knowledge transfer, while protecting the data privacy.
Biosketch of the speaker: Yaochu Jin is Chair Professor for AI with the School of Engineering, Westlake University, Hangzhou, China. He was an Alexander von Humboldt Professor for Artificial Intelligence, with the Faculty of Technology, Bielefeld University, Germany. Prior to that, he was a Surrey Distinguished Chair, Professor in Computational Intelligence, Department of Computer Science, University of Surrey, Guildford, U.K. He was a “Finland Distinguished Professor” of University of Jyväskylä, Finland, “Changjiang Distinguished Visiting Professor”, Northeastern University, China, and “Distinguished Visiting Scholar”, University of Technology Sydney, Australia. His main research interests include evolutionary optimization and learning, trustworthy machine learning and optimization, and evolutionary developmental AI. He is a Member of Academia Europaea and Fellow of IEEE.
Prof Jin is presently the President of the IEEE Computational Intelligence Society and Editor-in-Chief of Complex & Intelligent Systems. He was the Editor-in-Chief of the IEEE Transactions on Cognitive and Developmental Systems. He is the recipient of the 2018, 2021 and 2024 IEEE Transactions on Evolutionary Computation Outstanding Paper Award, and the 2015, 2017, and 2020 IEEE Computational Intelligence Magazine Outstanding Paper Award.
Abstract: Explainable artificial intelligence (XAI) has recently attracted great attention due to its importance in critical application domains such as self-driving cars, law, and healthcare. Genetic programming (GP) is a powerful evolutionary algorithm for AI, machine learning and AutoML. Compared with other standard machine learning models such as neural networks, the models evolved by GP tend to be more interpretable due to the symbolic nature of the model structure. GP model interpretability has been playing an increasingly important role in XAI. This talk will first briefly overview explainable AI, then focus on how GP helps XAI in three aspects: intrinsic XGP, aiming to directly evolve more interpretable (and effective) models by GP; post-hoc XGP, which uses GP to explain other black-box machine learning models and the models evolved by GP by simpler models (e.g. linear models); and GP visualisation, which focuses on visual interpretability. Finally, I will discuss the challenges and future directions for XGP.
Biosketch of the speaker: Mengjie Zhang is a Fellow of Royal Society of New Zealand, a Fellow of Engineering New Zealand, a Fellow of IEEE, an IEEE Distinguished Lecturer, Professor of Computer Science (Artificial Intelligence) at Victoria University of Wellington, where he heads the interdisciplinary Evolutionary Computation and Machine Learning Research Group. He is also the founding Director of the Centre for Data Science and Artificial Intelligence at the University.
His research is mainly focused on AI, machine learning and big data, particularly in evolutionary learning and optimisation, feature selection/construction and big dimensionality reduction, computer vision and image analysis, scheduling and combinatorial optimisation, classification with unbalanced data and missing data, and evolutionary deep learning and transfer learning. Prof Zhang has published over 900 research papers in refereed international journals and conferences. He has been serving as an associated editor for over ten international journals including IEEE Transactions on Evolutionary Computation, IEEE Transactions on Cybernetics, the Evolutionary Computation Journal (MIT Press), and involving major AI and EC conferences as a chair. He received the “Evo* Award for Outstanding Contribution to Evolutionary Computation in Europe 2023”. Since 2007, he has been listed as a top five (currently No. 3) world genetic programming researchers by the GP bibliography (http://www.cs.bham.ac.uk/~wbl/biblio/gp-html/index.html).
Prof Zhang is currently the Chair for IEEE CIS Awards Committee. He is also a past Chair of the IEEE CIS Intelligent Systems Applications Technical Committee, the Emergent Technologies Technical Committee and the Evolutionary Computation Technical Committee, the Chair for IEEE CIS PubsCom Strategic Planning subcommittee, and the founding chair of the IEEE Computational Intelligence Chapter in New Zealand.
Abstract: Evolutionary computation is a kind of optimization method inspired by the nature selection mechanism, “survival of fittest”, and has been widely used in a variety of fields. This talk will give a brief introduction of evolutionary computation, and then focus on introducing how to use evolutionary computation the solve the problems on complex networks, especially on two problems, namely, community detection and network robustness optimization.
Biosketch of the speaker: Jing Liu received the B.S. degree in computer science and technology and the Ph.D. degree in circuits and systems from Xidian University in 2000 and 2004, respectively. In 2005, she joined Xidian University as a lecture, and was promoted to a full professor in 2009. From Apr. 2007 to Apr. 2008, she worked at The University of Queensland, Australia as a postdoctoral research fellow, and from Jul. 2009 to Jul. 2011, she worked at The University of New South Wales at the Australian Defence Force Academy as a research associate. Now, she is the deputy dean and full professor in the Guangzhou Institute of Technology, Xidian University. Her research interests include evolutionary computation, complex networks, fuzzy cognitive maps, and computer vision. Prof. Liu has been the associate editor of IEEE Trans. Evolutionary Computation from 2015-2020, and the chair of Emerging Technologies Technical Committee (ETTC) of IEEE Computational Intelligence Society from 2017-2018.