Seminars
The Mathexlab seminar series is a set of seminars in the intersection of Explainable AI, Complex Systems and Physics. Distinguished speakers are invited to share insights with respect to the following themes:
1. High-Fidelity Multi-Physics Simulation Tools: This track addresses the development and application of knowledge-based (i.e., equation-based) computational tools for simulating real-world phenomena, combining multiple physical models to achieve high accuracy in fields like aerospace and robotics and computer graphics.
2. Analysis of Complex Systems through Dynamical Systems Theory: This track covers the use of dynamical systems theory to identify and analyze patterns in complex, time-evolving systems, focusing on theoretical frameworks and computational methods for understanding behavior in fields such as climate systems and weather.
3. AI + Domain Knowledge: This track explores how AI integrates with and can enhance specific domain expertise, investigating the role of AI in augmenting, automating, and expanding decision-making processes across various fields such as medicine, finance, and engineering.
4. Explainable AI (Theoretical and Applied): This track explores both the foundational theory and practical implementation of explainable artificial intelligence, focusing on methods that enhance transparency, interpretability, and accountability of AI systems, ensuring that models’ decisions are understandable to humans.
Biography: Spencer Sherwin is Head of Department and Professor of Computational Fluid Mechanics in the Department of Aeronautics at Imperial College London. He received his MSE and PhD from the Department of Mechanical and Aerospace Engineering Department at Princeton University. Prior to this he received his BEng from the Department of Aeronautics at Imperial College London.
Abstract: Advanced high order methods using Spectral/hp element discretization [1] including Galerkin, Discontinuous Galerkin (DG) and Flux Reconstruction (FR) formulations are gaining notable interest in both the academic and industrial sectors. The compact nature of the approach is not only attractive from the perspective of implementation on modern computational hardware but also provides a consistent geometric and spatially localized accuracy unlike many high order finite volume methods. These features make the methodology attractive in complex geometry flows involving transitional and turbulent boundary layers demanding a high level of accuracy for high end engineering applications that commonly arise in the automotive and aeronautical sectors.
Coming soon!