gradient-flow

Discrete approximation to 3D winding number mapping T^3->U(N)

[NeurIPS'25] Sequence Modeling with Spectral Mean Flows, in PyTorch

High-Dimensional Analysis of Gradient Flow for the Matrix Factorization Problem

Clean-room PyTorch reproduction of learning Wasserstein gradient flows from snapshot data via particle stitching and Wasserstein residuals (arXiv:2607.04738).

Supplementary code for the paper 'Are GATs Out of Balance?' to be published at NeurIPS 2023

Unity Equilibrium Theory - Dual-field gradient flow simulator for phase dynamics

Utilities of quantum theory of (higher-form) general gauge fields

A continuous-time geometric formulation of sorting as gradient flow on the permutohedron.

Discreteness as the Late-Time Limit of a Smooth Flow

Cobebase for my paper "Convergence of Shallow ReLU Networks on Weakly Interacting Data", accepted at NeurIPS 2025

Learning as optimization & dynamical systems, autodiff, backprop, gradient flow from scratch.

Analysis of the anisotropic Gradient Flow output of 1203.4469 for anisotropic lattice QCD gauge anisotropy determination and scale setting

A detailed-balanced chemical reaction network realized as a gradient flow, where equilibrium propagation is rigorous — the chemistry computes its own weight update (verified: local update == true gradient). Audits 'who computes the update?' across a co-location spectrum vs an offline ceiling. Part of physical-learning-substrates.

Kinetic Adaptation via Wasserstein Heuristics and Identity

From-scratch (pure NumPy) comparison of Explicit, Implicit & Newton gradient descent as discretizations of gradient flow.

Lean 4 / Mathlib formalization of finite-N Kuramoto synchronization: ODE existence + uniqueness, order-parameter bounds, Lyapunov descent, and convergence to synchrony for all-to-all AND general symmetric coupling with a uniform positive off-diagonal floor. Sorry-free, axiom footprint {propext, Classical.choice, Quot.sound}.

MILC collaboration's fork of the Quantum EXpressions lattice field theory framework

Accelerated Stein Variational Gradient Descent for sampling for densities

Independent research on inactive-neuron entry-exit dynamics, with mathematical analysis, archived experiments, reproducibility guides, and a connected research roadmap.