2023 IEEE Computational Intelligence Society Shenzhen Chapter Spring Symposium
07 April 2023 (Friday)
14:00 – 17:15 PM (Beijing Time, GMT+8)
Online via Tencent Meeting
Meeting ID: 408-496-064
Organized by the IEEE Computational Intelligence Society Shenzhen Chapter.
Agenda of the Activity
14:00-14:05
Opening ceremony
Associate Professor Ran Cheng, Southern University of Science and Technology
14:05-14:30
Keynote Speech: Pareto Set Learning and its Applications
Dr. Xi Lin, City University of Hong Kong
14:30-14:55
Keynote Speech: Safe & Efficient Machine Learning
Dr. Liangli Zhen, Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), Singapore
14:55-15:20
Keynote Speech: Multi-Objective Archiving
Assistant Professor Miqing Li, University of Birmingham
15:20-15:45
Keynote Speech: When Computational Intelligence Meets Biosystems
Assistant Professor Shouyong Jiang, University of Aberdeen, Scotland
15:45-16:00
Coffee break
16:00-16:25
Keynote Speech: Towards Next-Generation Auditory Information Processing: A Neuromorphic Approach
Assistant Professor Jibin Wu, Hong Kong Polytechnic University
16:25-16:50
Keynote Speech: Evolutionary Deep Learning for Image Analysis
Professor Bing Xue, Victoria University of Wellington
16:50-17:15
Keynote Speech: Genetic Programming vs Reinforcement Learning: A Case Study on Job Shop Scheduling
Associate Professor Yi Mei, Victoria University of Wellington
17:15
Closing ceremony
Talks & Speakers
1. Pareto Set Learning and its Applications
Abstract:
Multi-objective optimization problems can be found in many machine learning applications, such as multi-task learning, multi-objective Bayesian optimization, and neural multi-objective combinatorial optimization. These problems have multiple objectives to optimize, and no single solution can optimize all the objectives simultaneously. Different multi-objective optimization algorithms have been proposed to find a finite set of Pareto solutions with different optimal trade-offs among the objectives. However, it is well-known that the whole Pareto set is on a continuous manifold and can contain infinite solutions. The structural properties of the Pareto set are not well exploited in existing methods, and the finite-set approximation may not contain the most preferred solution(s) for decision-makers. In this talk, we will discuss how to develop a novel learning-based method to approximate the whole Pareto set for a given multi-objective optimization problem, which generalizes the decomposition-based multi-objective optimization algorithm (MOEA/D) from finite populations to models. With our proposed Pareto set model, decision-makers can easily explore any trade-off area in the approximate Pareto set for flexible decision-making. We will also present our current work on Pareto set learning for multi-objective Bayesian optimization and neural multi-objective combinatorial optimization.
Dr. Xi Lin, is currently a Post Doctoral Research Fellow with the Department of Computer Science at the City University of Hong Kong. He received the Ph.D. degree in 2020 from City University of Hong Kong under the supervision of Prof. Qingfu Zhang. His research interests include multi-objective optimization, multi-task learning, Bayesian optimization, evolutionary computation, and learning for optimization. His work has been published in top-tier machine learning conferences such as Conference on Neural Information Processing Systems (NeurIPS) and International Conference on Learning Representations (ICLR). He is a regular reviewer for top-tier machine learning and evolutionary computation conferences/journals, such as NeurIPS, ICML, ICLR, JMLR, TEVC and TCYB, and has received multiple top reviewer awards from ICML and ICLR.
2. Safe & Efficient Machine Learning
Abstract:
The remarkable success of Artificial Intelligence (AI) across various domains comes with significant safety and security risks, such as vulnerability to adversarial attacks and sensitivity to environmental changes, particularly in critical fields like healthcare, surveillance, and public security. Furthermore, state-of-the-art machine learning models often demand extensive data, computing resources, and human resources, which can be costly and limited in numerous applications. In this talk, Dr. Zhen will discuss the challenges and solutions to AI safety and efficiency in real-world applications. He will introduce his team’s research endeavors in addressing these challenges, beginning with the motivation for safe and efficient AI research. He will then present some of their work in these research areas and conclude by introducing the grants and projects that support these research activities.
Dr. Liangli Zhen, is a Research Scientist at the Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), Singapore. He also serves as the Group Manager at IHPC, where he leads a team dedicated to enhancing the robustness and efficiency of AI solutions for real-world applications. The outcomes of his research have been published in leading conferences and journals such as ICLR, CVPR, SIGIR, IEEE TPAMI, IEEE TNNLS, and IEEE CIM, and have demonstrated successful implementation in real-world systems. His works won the Sichuan Province Computer Federation Outstanding Paper Award in 2018 and the Best Student Paper Award of the International Conference on Simulated Evolution and Learning in 2017.
3. Multi-Objective Archiving
Abstract:
Two pivotal issues centred around evolutionary multiobjective optimisation are solution generation and population maintenance. The first one is concerned with mating selection and variation (e.g., crossover and mutation) where we want to generate promising offspring, hopefully better than their parents. The other is concerned with environmental selection (aka elitism) where we want to avoid losing the very best solutions found during the search course. The second issue can also be generalised as archiving, a process of taking new solutions, comparing them with the old ones, and deciding how to update the population/archive. In this tack, I will briefly introduce archiving, including its importance, significance, problems of existing archiving methods, and potential research directions.
Dr. Miqing Li, is an Assistant Professor at the University of Birmingham and a Turing Fellow of the Alan Turing Institute, UK. Miqing’s research focuses on multi-objective optimisation, where he develops population-based randomised algorithms (mainly evolutionary algorithms) for both basic and practical problems. On basic research, Miqing tackles general challenging problems (e.g. problems with many objectives, complex constraints, or expensive to evaluate). These works were published in top journals in the EC areas such as TEVC, ECJ, TELO, TCYB and IEEE CIM, some of which are among widely used ones in the field. On applied research, working with experts in different fields (e.g., software engineering, high-performance computing, neural architecture search, and supply chain), Miqing develops novel methods to tackle their practical problems. These works appear in high-profile venues in relevant areas such as TOSEM, TSE, ICSE, FSE, AIJ, NeurIPS, TPDS, TFS, TDSC, TAAS, TSMC-A, TR, EJOR, and CSUR. Some of these methods have become the most cited works in the corresponding venues since they made appearance (e.g., TPDS 2016 and TOSEM 2016). His work has received the Best Paper Award/nomination in mainstream conferences in the EC field (CEC, GECCO, and SEAL). Miqing was the founding chair of the IEEE CIS’ Task Force on Many-Objective Optimisation, and is an Associate Editor of IEEE Transactions on Evolutionary Computation.
4. When Computational Intelligence Meets Biosystems
Abstract:
Computational intelligence has been an important alternative to traditional operations research approaches for decision making in complex systems. In this talk I am going to first present our recent work on the development of fast evolutionary algorithms with dimensionality reduction to handle dynamic multi-objective optimisation. Then I will discuss some applications of computation intelligence to improve our understanding of biological functions of biosystems and offer new designs of such systems for industrial biomanufacturing. Such challenges of computational intelligence will also be highlighted from an applied perspective.
Dr. Shouyong Jiang, is an assistant professor at the University of Aberdeen, Scotland. Prior to that, he was a lecturer at the University of Lincoln, and a postdoc research fellow at Newcastle University where he focused on computational intelligence driven biotechnology. He was a visiting researcher at University of Vigo (Spain) and ETH Zurich (Swiss). Dr Jiang is a fellow of higher education academy. He was a recipient of IEEE Computational Intelligence Society Outstanding PhD Dissertation Award, and CSC Excellent Student Studying Abroad Award. He has a track record of seven major research projects, as principal investigator/co-investigator, funded by Scotland, UK government, and EU North Sea Programme, with a total of over £10M in value. He has published ~50 research papers in computational intelligence and interdisciplinary fields. He is an editorial board member of several journals and a guest editor of special issues at CAAI Transactions on Intelligence Technology, Metabolites.
5. Towards Next-Generation Auditory Information Processing: A Neuromorphic Approach
Abstract:
Hearing is a vivid part of our conscious lives. Driven by recent advances in artificial intelligence (AI), the capability of machine hearing systems has improved by leaps and bounds over the last decade. The increased capability also comes along with challenges to efficiently, rapidly, and reliably process sound signals. With an ever-growing demand for human-computer auditory interfaces, these challenges are expected to be exacerbated. Throughout the history of machine hearing research, neuroscience has played a pivotal role by offering a bountiful source of inspiration for novel acoustic features, algorithms, and computational models. To achieve more capable, efficient, and reliable machine hearing systems, it calls for continuously exchanging ideas between the fields of neuroscience and AI.
By harnessing the findings and insights from neuroscience studies, the interdisciplinary neuromorphic computing research offers immense opportunities for building brain-inspired auditory systems for machine hearing. In this talk, I will present our recent research outcomes on neuromorphic auditory information processing that is grounded on the brain-inspired spiking neural networks (SNNs). In particular, I will present the technology breakthroughs in: 1) auditory neural codes that can efficiently and effectively encode sound signals; 2) learning algorithms that support rapid and efficient pattern recognition using SNNs; 3) task-specific neural architectures that grounded on the pre-existing neural structures; I will conclude the talk by sharing my long-term vision of developing low-power, reliable, adaptive, and explainable neuromorphic cognitive machines.
Dr. Jibin Wu, is currently an Assistant Professor in the Department of Computing, Hong Kong Polytechnic University. Before joining PolyU, he was a research scientist at Sea AI Lab (SAIL) from 2021 to 2022. Dr. Wu received Bachelor degree in Electrical Engineering and Ph.D. degree from National University of Singapore (NUS) in 2016 and 2020, respectively. His research interests broadly include brain-inspired artificial intelligence, neuromorphic computing, computational audition, speech processing, and machine learning. Particularly, he is dedicated to understanding the computational principles and architectures of biological brains, and to designing next-generation brain-inspired cognitive machines that are intelligent, low-power, robust, adaptive, and explainable. Dr. Wu has actively published in prestigious conferences and journals in artificial intelligence and speech processing, including NeurIPS, AAAI, TPAMI, TNNLS, TASLP, Neurocomputing, and IEEE JSTSP. He is currently serving as the Associate Editor for IEEE Transactions on Neural Networks and Learning Systems.
6. Evolutionary Deep Learning for Image Analysis
Abstract:
Image analysis is a fundamental task in a wide range of real-world problems. Deep learning, particularly deep neural networks (DNNs), have been a successful approach to image analysis, but designing an effective DNNs is extremely hard, requiring extensive experience and expertise in both DNNs and the problem domain as well as a huge computational cost. To address these limitations, evolutionary computation techniques start playing a significant role for automatically determining deep structures to tackle image classification tasks, and have great potential to advance the developments of deep structures and algorithms. This talk will provide an extended view of deep learning, overview the state-of-the-art work in evolutionary deep learning. Furthermore, we will discuss some recent developments using Genetic Programming (GP) to automatically evolving deep structures and feature learning for image analysis with a highlight of the interpretation capability and visualisation of the constructed features.
Dr. Bing Xue, is currently Professor of Artificial Intelligence, and Deputy Head of School for Engineering and Computer Science at Victoria University of Wellington (VUW). Her research focuses mainly on evolutionary computation and machine learning, such as feature selection, evolutionary deep learning, and image analysis, and their real-world applications in biology, healthcare, aquaculture, forest, and others. She has over fully refereed 350 publications and leading several prestigious research grants.
Professor Xue is currently the Editor of IEEE CIS Newsletter, Chair of IEEE CIS Evolutionary Computation Technical Committee, and Chair of IEEE CIS Task Force on Evolutionary Deep Learning and Applications. She also chairs many international conferences, such as General Chair of PRICAI 2025, Conference chair of IEEE CEC 2024, Proceedings chair of ACM GECCO 2023, and Panel chair of CEC 2023. She has also served as an Associate Editor of several international journals, such as IEEE TEVC, TAI, CIM, TETCI, and ACM TELO. She is also a Fellow of Engineering New Zealand.
7. Genetic Programming vs Reinforcement Learning: A Case Study on Job Shop Scheduling
Abstract:
Reinforcement learning is an important branch in machine learning, which aims to learn policies for agents to make decisions by interacting with the environment. Its popularity increases rapidly in recent years, and one of its interesting new applications is “learning to optimise” — it learns how to search in the solution space to solve an optimisation problem more effectively.
In this talk, we focus on a classic combinatorial optimisation problem: job shop scheduling, particularly in dynamic environment where unpredicted new jobs arrive in real time and must be incorporated into the schedule on-the-fly. Dispatching rules have been commonly used to make real-time scheduling decisions for solving the dynamic job shop scheduling problem. Our group has been working on using genetic programming to automatically learn dispatching rules for many years. In recent years, people started to use reinforcement learning to learn dispatching rules as policies. Thus, it would be interesting to compare genetic programming and reinforcement learning in this case. In this talk, we will briefly introduce how to learn dispatching rules for dynamic job shop scheduling by genetic programming and reinforcement learning, and then discuss about the pros and cons of both techniques. We hope this talk to give some insight on how to learn better policies, especially for complex combinatorial optimisation problems.
Dr. Yi Mei, is an Associate Professor and Postgraduate Coordinator (Science) at the School of Engineering and Computer Science, Victoria University of Wellington, Wellington, New Zealand. He received his BSc and PhD degrees from University of Science and Technology of China in 2005 and 2010, respectively. His research interests include evolutionary computation for combinatorial optimisation, genetic programming, automatic algorithm design, explainable AI, multi-objective optimisation, transfer/multitask learning and optimisation.
Yi has ~200 fully refereed publications, including the top journals in EC and Operations Research (OR) such as IEEE TEVC, IEEE Transactions on Cybernetics, European Journal of Operational Research, ACM Transactions on Mathematical Software, and top EC conferences (GECCO). He won an IEEE Transactions on Evolutionary Computation Outstanding Paper Award 2017, a GECCO Best Paper Award 2022 and the EuroGP Best Paper Award 2022. He is an Associate Editor of IEEE Transactions on Evolutionary Computation, an Editorial Board Member of four other international journals. He serves as a Vice-Chair of the IEEE CIS Emergent Technologies Technical Committee, and a member of three IEEE CIS Task Forces and two IEEE CIS Technical Committees. He is the Chair of the IEEE Taskforce on Evolutionary Scheduling and Combinatorial Optimisation and the Chair of New Zealand Central Section. He is a Fellow of Engineering New Zealand and an IEEE Senior Member.