Introduction
This bi-monthly seminar series explores real-world applications of physics-informed machine learning (Φ-ML) methods to the engineering practice. They cover a wide range of topics, offering a cross-sectional view of the state of the art on Φ-ML research, worldwide.
Participants have the opportunity to hear from leading researchers and learn about the latest developments in this emerging field. These seminars also offer the chance to identify and spark collaboration opportunities.
About the event
CAUSALITY analysis is an important problem lying at the heart of science (e.g., Einstein 1953). The recent rush in artificial intelligence (AI) has stimulated enormous interest in causal inference, in expectation of developing a causal AI to overcome the interpretability crisis. Historically causal inference has been formulated as a statistical problem in data science; see the classics by the Nobel Prize and Turing Award winners such as Clive Granger, Judea Pearl, Guido Imbens, Joshua Angrist, among others. On the other hand, we have found that causality in terms of information flow/transfer is actually “a real notion in physics that can be derived ab initio” (Liang, 2016), rather than axiomatically proposed as an ansatz, and, moreover, can be quantified. A comprehensive study with generic systems has been fulfilled recently, with explicit formulas attained in closed form (Liang, 2008; 2016). An important corollary is that, in the linear limit, causation implies correlation, but correlation does not imply causation, expressing the long standing philosophical debate ever since Berkeley (1710) in a concise mathematical formula.
The above rigorous formalism has been validated with benchmark systems like baker transformation, Hénon map, and purportedly designed networks with nodes almost synchronized. They have also been applied to real world problems in the diverse disciplines such as climate science, meteorology, hydrology, turbulence, neuroscience, financial economics, quantum mechanics, etc., with interesting new findings. For example, Stips et al. (216) found that, while CO2 emission does drive the recent global warming, on a paleoclimate scale, it is global warming that drives the CO2 emission. In this talk, I will demonstrate, with realistic examples, how they can be employed to fulfill causality-aided scientific discovery in natural sciences, to design causal AI algorithm(s) for interpretability enhancement, and to apply the latter to develop intelligent systems for prediction and generalization.
Some References
Selected theoretical papers:
Liang, 2008: Information flow within stochastic dynamical systems. Phys. Rev. E, 78, 031113.
Liang, 2014: Unraveling the cause-effect relation between time series. Phys. Rev. E, 90, 052150.
Liang, 2016: Information flow and causality as rigorous notions ab initio. Phys. Rev. E, 94, 052201.
Liang, 2021: Normalized multivariate time series causality analysis and causal graph reconstruction. Entropy, 23, 679.
Selected applications:
Stips et al., 2016: On the causal structure between CO2 and global temperature. Sci. Rep. 6:21691.
Liang et al., 2021: El Niño Modoki can be mostly predicted more than 10 years ahead of time. Sci. Rep. 11:17860.
Cong et al., 2023: Altered default mode network causal connectivity patterns in autism spectrum disorder revealed by Liang information flow analysis. Human Brain Mapping DOI: 10.1002/hbm.26209.
Yi and Bose 2022: Quantum Liang information flow as causation quantifier. Phys. Rev. Lett., 129, 020501.
