Vakonomic Fluids
arXiv:2607.18312
2026
Dynamics
1 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper’s transferable asset is a variational discretization of constrained dynamics that retains Lie–Poisson structure rather than enforcing constraints only approximately at each step. Its discrete trajectories preserve Casimir invariants such as $\operatorname{tr}(\mathbf{Z}^{k})$ to machine precision, while a momentum-map/Clebsch representation provides a low-rank implementation. This suggests a structure-preserving optimizer or recurrent state update in which an auxiliary matrix state evolves by conjugation, preventing numerical drift in spectral invariants during long training or iterative inference. The most realistic first use is as a replacement for unconstrained momentum dynamics in a small matrix-parameterized optimizer, with ablations against AdamW and ordinary momentum.
Ideas from this paper
Unverified
2026
Introduce an auxiliary matrix-valued optimizer state whose update is a Lie–Poisson flow discretized by similarity transforms rather than additive Euler steps. Because similarity transforms preserve $\operatorname{tr}(Z^k)$ and the full eigenvalue multiset, long training runs avoid spectral drift in the optimizer state; the state can then generate a preconditioned update for ordinary neural-network parameters.
Useful5/10
Difficulty6/10
Novelty7/10