Ideas, with the knobs exposed
Writing
Interactive essays on scientific computing, notes from the research, and things I wanted to understand better.
What symmetry buys a learned turbulence model
What do symmetry constraints buy a neural turbulence closure? Same accuracy floor, 25× fewer parameters, and one missing input for Reynolds-number generalization. An interactive companion to the paper.
Interactive essayYour time integrator is a filter
Time integration acts as a filter too: at practical CFL numbers the time-integration error dominates the coarse-simulation residual, and one extra closure term fixes it. Interactive companion to the paper.
Interactive essayYour grid is a filter
An interactive tour of exact expressions for the unresolved stress in finite-volume LES: the grid is a filter, and the classic subgrid stress is the wrong modeling target.
Interactive essayMake Beamer slides 16:9
How to make LaTeX Beamer slides widescreen 16:9 instead of the default 4:3.
NoteWriting a differentiable fluid solver in Julia
A journey from sparse matrices in MATLAB to differentiable GPU operators in Julia, building IncompressibleNavierStokes.jl.
DraftWhat is the difference between forward and reverse mode automatic differentiation?
Forward and reverse mode automatic differentiation compute the same chain of matrix products in different orders — and that order determines their cost.
Note