Modeling and Simulation Group Meeting
Two talks about Information Geometric Regularization
Speaker: Ed Chen, Alex Albors, Courant, NYU
Location: Warren Weaver Hall 1314
Date: Thursday, November 12, 2026, 12:30 p.m.
Synopsis:
Ed Chen: Toward Fully Explicit Information Geometric Regularization
Information Geometric Regularization (IGR) introduces an inviscid PDE-based shock regularization for high speed flows, which requires solving an auxiliary screened-Poisson elliptic equation to compute a regularizing artificial entropic pressure. We aim to simplify IGR by introducing a time evolution equation for the entropic pressure, enabling its explicit time integration and implementation in standard hyperbolic conservation law packages.
Alex Albors: Information Geometric Regularization adjoints for Shock-laden Aerodynamic Shape Optimization
Shocks are ubiquitous in compressible-flow simulations and strongly influence aerodynamic forces, stability, and design. Information-geometric regularization (IGR) is an inviscid, PDE-level regularization of shocks in compressible flow, smoothing discontinuities without dissipating oscillatory structures or introducing nondifferentiable limiters. We correct the treatment of curved solid walls in IGR regularization. This correction is necessary for a well-posed adjoint formulation and enables sensitivity analysis in the presence of shocks and complex geometries, unlike conventional shock-capturing schemes. It also restores the proper interaction between IGR and curved-wall boundary conditions in the forward solve. We then discuss adjoint consistency with IGR shock capturing for the compressible Euler equations, deriving both continuous and discrete adjoint formulations. The resulting framework supports accurate gradient computation for aerodynamic shape optimization and shock control.