Title: Predicting critical transition and system collapse with machine learning
Abstract: To predict a critical transition due to parameter drift based on data is an outstanding problem in complex dynamics and applied fields. A closely related problem is to predict whether the system is already in or if the system will be in a transient state preceding its collapse. A model free, machine-learning solution to both problems will be presented. The idea is to develop a parameter-cognizant machine-learning framework based on reservoir computing. When the machine is trained only with data from the normal functioning regime with oscillatory dynamics (i.e., before the critical transition), the transition point can be predicted accurately. Remarkably, for a parameter drift through the critical point, the machine is able to predict not only that the system will be in a transient state, but also the average transient time before the final collapse. Applications to electrical power and ecological systems will be demonstrated. The machine-learning framework can also be extended to predicting synchronization transition and amplitude death in coupled nonlinear systems.
Speaker Bio: Ying-Cheng Lai is the ISS Endowed Professor of Electrical Engineering and a Professor of Physics at Arizona State University. He was a PECASE recipient in 1997 and has been a Fellow of the American Physical Society since 1999. In 2016, he was selected by the Pentagon for the Vannevar Bush Faculty Fellowship. In 2018, he was elected as a Foreign Member of National Academy of Science and Letters of Scotland. In 2020, he was elected as a Foreign Member of Academia Europaea (The Academy of Europe) and as a Fellow of the American Association for the Advancement of Science (AAAS). As of March 2021, Y.-C. Lai has published over 500 refereed-journal papers with more than 25000 citations (H-index: 75; i-10 index: 388).
Title: Synergetic Interplay between Artificial Intelligence and Complex Network
Abstract: Artificial Intelligence (AI) as an enabling intelligent systems technology is playing a more and more important role in today’s industry and society. Complex Networks (CN), on the other hand, represent characteristics of many large-scale real-world network systems. The recent advances in AI have provided a powerful platform technology to solve complex problems, while CN presents an alternative way to make problem-solving simpler and faster.
In this talk, we will first discuss recent developments in both AI and CN, and then examine emerging issues associated with synergetic interplay between them to bring out the best of both fields. We will also touch on potential new thinking paradigms beyond AI to deal with complex problems arising from these systems, speculating innovative methodologies inspired by the Nature for the future.Particular attention will be given to the modelling, control and optimisation issues in large-scale industrial engineering systems such as smart grids. Several real-world industrial problems including some of our own research work will be used as case studies.
Speaker Bio: Distinguished Professor Xinghuo Yu is an Associate Deputy Vice-Chancellor and a Vice-Chancellor’s Professorial Fellow at RMIT University (Royal Melbourne Institute of Technology), Melbourne, Australia. He is also the Junior Past President of IEEE Industrial Electronics Society. His main research areas include control systems engineering, intelligent and complex systems, and future energy systems. He received many awards and honours for his contributions, including the 2018 MA Sargent Medal from Engineers Australia, the 2018 Australasian AI Distinguished Research Contribution Award from Australian Computer Society, and the 2013 Dr.-Ing. Eugene Mittelmann Achievement Award from IEEE Industrial Electronics Society. He was one of the 15 Shortlist Nominees for the 2020 Global Energy Prize and named a Highly Cited Researcher by Clarivate Analytics in 2015-2020. He is a Fellow of IEEE, Engineers Australia, Australian Computer Society, and Australian Institute of Company Director.
Speaker Bio:周志华，南京大学计算机系主任兼人工智能学院院长、校学术委员会委员，主要从事机器学习与人工智能研究，在集成学习、多标记学习与弱监督学习方面有重要贡献。著有《机器学习》《Ensemble Methods: Foundations and Algorithms》等，论著被引用5万余次，成果在华为等企业转化实施，并服务于国家重大工程。获国家自然科学二等奖、3次教育部自然科学一等奖、IEEE计算机学会Edward J. McCluskey技术成就奖、CCF王选奖等，是欧洲科学院外籍院士，ACM、AAAI、IEEE等的Fellow。
Title:The ``endo-exo'' problem in complex systems (ecology, earthquakes, financial volatility, epileptic seizures…)
Abstract:The endo-exo problem-- i.e., decomposing system activity into exogenous and endogenous parts -- lies at the heart of statistical identification in many fields of science. E.g., consider the problem of determining if an earthquake is a mainshock or aftershock, or if a surge in the popularity of a youtube video is because it is going viral, or simply due to high activity across the platform. The endo-exo problem is also at the heart of a general description of the dynamics of out-of-equilibrium complex systems generalising the fluctuation-susceptibility theorem. I will present recent exciting results obtained in my group, which include
1) the development of a powerful Expectation Maximization (EM) algorithm and objective statistical criteria (BIC) to select the flexibility of the deterministic background intensity of self-exciting Hawked point processes that have enjoyed great recent popularity and rapid development, with application to determine the Soros level of reflexivity in finance;
2) an augmented Epidemic-Type Aftershock Sequence (ETAS) model that accounts for the spatial variability of the background rates, anddirect quantitative test of criticality of the Earth crust;
3) the mapping of the non-Markovian Hawkes self-excited point process (whichprovides an efficient representation of the bursty intermittent dynamics of many physical, biological, geological, and economic systems) onto stochastic partial differential equations that are Markovian and the development of newfield theoretical approach in terms of probability density functionals, with novel results;
4) the discovery that a wide class of nonlinear Hawkes processes have the PDF of their intensities described by Zipf’s law. These methods and results are relevant to many scientific fields from linguistic, social, economic, computer sciences to essentially all natural sciences. In parallel, self-excited dynamics is a prevalent characteristic of many systems, from the physics of shot noise and intermittent processes, seismicity, financial volatility and financial defaults, to sociology, consumer behaviors, computer sciences, The Internet, neuronal discharges and spike trains in biological neuron networks, gene expression and even criminology.
Speaker Bio:Prof. Didier Sornette is a member of the Academia Europaea, a member of the Swiss Academy of Engineering Sciences (SATW), Dean and Chair Professor at the Institute of Risk Analysis, Prediction and Management (Risks-X) at the Southern University of Science and Technology(SUSTech), a full professor on the Chair of Entrepreneurial Risks at ETH Zurich, and also associated to Department of Earth Science and Department of Physics at ETH Zurich. Moreover, he is the director of the Financial Crisis Observatory, co-founder of the ETH Risk Center, and professor of finance at the Swiss Finance Institute. He is a fellow of the American Association for the Advancement of Science (AAAS) and a fellow of the World Innovation Foundation (WIF).
Prof. Sornette is a world-class expert in the field of complex systems and extreme risk management. He developed the Dragon King extreme event theory, which uses rigorous data-driven mathematical and statistical analysis methods to identify, control and predict complex system instabilities and extreme risks, with successful applications in a range of complex systems including financial risks, earthquake prediction, nuclear energy security, cyber-security, social networks, health systems, etc. He founded the Financial Crisis Observatory(FCO), monitoring over 20,000 different financial assets globally in real time, and has successfully predicted many market turbulences including the bursting of three Chinese stock market bubbles in 2007, 2009 and 2015, the 2008 crude oil bubble, and the decoupling of the EUR-CHF in 2011. He has published over 800 journal papers and 10 books, with 47,000 Google Scholar citations and an H-index of 105. Meanwhile, he has also held roles as an expert advisor to several world-renowned aerospace companies, banks, funds and reinsurance companies, such as the Chief Risk Advisor at Bank of America, an external expert at Los Alamos National Laboratories and member of the Board of the “Fondation d’entreprise SCOR pour la Science”.
Speaker Bio: 王正明，国防科技大学教授，“国家百千万工程”一、二层次人选，全国优秀博士学位论文作者，系统科学、数学专业博士生导师，享受政府特殊津贴。长期从事装备试验设计、评估与鉴定等领域的科研和教学工作。现任军队数学教学联席会主席，国防科技大学教学委员会主任、学术委员会副主任。获军队科技进步一等奖3项，二等奖4项，国家教学成果二等奖1项，军队教学成果一等奖3项。主持国家及军队级项目20余项。合作发表论文180篇，出版5部共约300万字的专著，获解放军图书奖1项。立二等功1次。