Memory with Onsager-Casimir symmetry: Rotating particle in a viscoelastic fluid

arXiv:2608.00344 2026 Dynamics 2 ideas extracted · analyzed Aug 31, 2026

What the math gives to ML

The paper provides a transferable mechanism for combining non-Markovian memory with a time-antisymmetric rotational coupling. Eliminating a slow bath variable produces a non-reciprocal memory kernel whose orientation rotates in time, while Onsager–Casimir symmetry relates the kernel and response under reversal of the rotation vector. For neural networks, this suggests recurrent or state-space modules with a stable decaying symmetric memory component plus a skew-symmetric rotating component, rather than unconstrained recurrent matrices. The paper also supplies a fluctuation–response identity that can become a measurable training regularizer linking hidden-state cross-correlations to impulse responses.

Ideas from this paper

✓✓ Beats tuned baseline 2026

Stable Rotating-Memory State Space

Replace an unconstrained recurrent transition with a decaying symmetric memory operator plus a skew-symmetric rotational operator. The skew component creates phase-shifted cross-channel memory and can represent oscillatory or circulatory temporal dependencies without requiring eigenvalues with large positive real parts.

Useful8/10
Difficulty5/10
Novelty6/10
Paper: Memory with Onsager-Casimir symmetry: Rotating particle in a viscoelastic fluid arXiv:2608.00344
Mechanism failed 2026

Onsager–Casimir Response Regularizer

Train a sequence model so that measured perturbation responses and spontaneous hidden-state correlations satisfy the paper's off-diagonal fluctuation–response identity. This discourages arbitrary non-reciprocal dynamics while preserving a controlled antisymmetric response that can encode directional temporal dependencies.

Useful7/10
Difficulty6/10
Novelty8/10
Paper: Memory with Onsager-Casimir symmetry: Rotating particle in a viscoelastic fluid arXiv:2608.00344