Machine learning for fluid dynamics
4-6 March 2026
CWI Amsterdam
CWI Amsterdam
| Home | Program | Photos | Venue/Dinner | Challenge |
| 4 March 10am | Causal Inference for Scientific Discovery in Fluid Dynamics |
| Adrian Lozano Duran (CalTech) | |
| Causality lies at the core of scientific discovery, yet uncovering it remains a fundamental challenge, particularly in modern fluid-dynamics datasets. These systems involve nonlinear interactions, feedback, and self-causation, as well as the intertwined roles of mediators, confounders, and colliders, often further obscured by unobserved exogenous influences. Here, we discuss the role of causal inference in fluid dynamics. We introduce a family of new information-theoretic approaches that measure causal influence as the incremental information about future events gained from past observations. We discuss three levels of detail: average causality, state-dependent causality, and space--time causality. The approaches are non-intrusive and require only paired past--future samples, making them practical for both computational and experimental studies, even when data are scarce. We apply the methods to challenging cases for causal inference, including the energy cascade in turbulence and inner--outer interactions in wall-bounded turbulence, and we outline extensions to fully 3D space--time data. | |
| 5 March 9am | Solution discovery in fluids with high precision using neural networks |
| Ching-Yao Lai (Stanford University) | |
| I will discuss examples utilizing neural networks (NNs) to find solutions to partial differential equations (PDEs) that facilitate new discoveries. Despite being deemed universal function approximators, neural networks, in practice, struggle to fit functions with sufficient accuracy for rigorous analysis. Here, we developed multi-stage neural networks that can reduce the prediction error to nearly the machine precision of double-precision floating points within a finite number of iterations. We use accurate NNs to tackle the challenge of searching for singularities in fluid equations ( Wang-Lai-Gómez-Serrano-Buckmaster, Phys. Rev. Lett. 2023 ). Unstable singularities, especially in dimensions greater than one, are exceptionally elusive. With NNs we demonstrate the first discovery of smooth unstable self-similar singularities to unforced incompressible fluid equations ( Wang et al., arXiv:2509.14185 ). The example illustrates how deep learning can be used to discover new and highly accurate numerical solutions to PDEs. | |
| 6 March 9am | Data over Dogma: Ruthless Empiricism, Strange Ideas, and the Future of Weather Forecasting |
| Daniel Worrall (Google Deepmind) | |
| Accurate, reliable weather forecasts are fundamental to protecting life and infrastructure across the globe. While the field has historically been dominated by numerical weather prediction and the simulation of discretised physics, we are currently witnessing the largest disruption to modern meteorology in decades. Machine learning based forecasting has evolved from experimental curiosity to operational reality in a matter of years, now delivering systems that rival and surpass the top physics-based models. This talk presents Google DeepMind’s latest weather forecasting system, discussing how our development process is guided by an uncompromising empiricism—a philosophy that prioritises statistical evidence over traditional physical intuition. This data-first approach has led to the adoption of strange and counter-intuitive architectural choices that, while unconventional to the fluid dynamics community, have proven essential for blazingly fast, accurate predictions. We will also explore the broader implications of this shift: how machine learning is not just accelerating the way we practice science, but fundamentally altering our understanding of how to model complex dynamical systems. | |
| 6 March 3:30pm | Is AI what's next for fluids? |
| Gianluca Iaccarino (Stanford University) | |
| While the public has been both fascinated and frightened by chatGPT and AGI, a (relatively) quiet revolution is brewing in the science community. The Nobel prize for chemistry awarded to the AlphaFold team and efforts to create AI scientists (Autoscience Institute) seem to indicate a clear path. The US government investments in AI and computing infrastructure with the concurrent reduction of budget for traditional funding agencies also appears to signal a changing environment. Today LRMs (large reasoning models) can easily pass exams in undergraduate and graduate fluids classes, but it is unclear if or how this objective competence translates into insight, knowledge and ultimately innovation. The keynote will provide a personal perspective on the role of inductive and deductive thinking in science. I will focus on the role that AI can have on accelerating the “classic" scientific process (observation-inquiry-hypothesis-experiment-analysis) and some of the current critical bottlenecks, especially in the realm of Agentic AI. Examples and differences in the development and success of AI for forecast, design and discovery will conclude the talk. |
