# Harmonic-Mode Branch for Topological Memory

- ID: 2562
- Canonical URL: https://synthcore.org/idea/2562/harmonic-mode-branch-for-topological-memory
- API JSON: https://synthcore.org/api/idea/2562.json
- API Markdown: https://synthcore.org/api/idea/2562.md
- Verification status: mech_ok_no_baseline
- Source: [arXiv:2608.19360](https://arxiv.org/abs/2608.19360)
- Category: architecture
- Solves: stability, memory, accuracy
- ML areas: graph-nn, ssm, rnn, memory
- Math tags: dynamical-systems, geometry, linear-algebra, spectral-theory, topological structure
- Ratings: usefulness 7/10; difficulty 6/10; novelty 7/10

## Idea description

Do not force Hodge dissipation onto harmonic edge modes, because these modes are precisely the obstruction to global coercivity. Split the latent state into dissipative coexact modes and a finite-dimensional harmonic branch, and use harmonic-decoupled interactions so each harmonic coordinate defines an invariant affine fibre with its own attractor.

## Mathematical statement

Let $L$ be the Hodge Laplacian restricted to the divergence-free edge space $V$, with orthogonal decomposition $V=\operatorname{range}(L)\oplus\ker L$. The dissipative component $z_\perp$ has spectral decay controlled by $\lambda_1^+=\min\{\lambda>0:\lambda\in\operatorname{spec}(L)\}$, whereas the harmonic component $h\in\ker L$ is not damped by $L$. Let $\Pi_H$ be the orthogonal projector onto $\ker L$. A harmonic-decoupled interaction satisfies $\Pi_HN(z_\perp,h)=0$ and may depend on $h$ parametrically. Consequently $h(t)=h(0)$, or it evolves only under a separately specified low-dimensional drift, and every affine set $h+\operatorname{range}(L)$ is invariant. This preserves topological degrees of freedom while retaining dissipative control in the orthogonal complement.

## Key formulas

- $$V=\operatorname{range}(L)\oplus\ker L,\qquad z=z_\perp+h,\quad z_\perp\in\operatorname{range}(L),\ h\in\ker L.$$
- $$Lh=0,\qquad \lambda_1^+=\min\{\lambda>0:\lambda\in\operatorname{spec}(L)\},\qquad \|z_\perp(t)\|\lesssim e^{-\nu\lambda_1^+t}\|z_\perp(0)\|.$$
- $$\Pi_HN(z_\perp,h)=0\Longrightarrow \dot h=0,\qquad h+\operatorname{range}(L)\ \text{is invariant}.$$

## Implementation notes

Use this decomposition in a graph neural network, mesh encoder, neural state-space model, or long-horizon recurrent predictor when the graph or cell complex has nontrivial first cohomology. Compute an orthonormal basis $H$ for $\ker L$ with an eigensolver, then form $\Pi_H=HH^\top$ and $\Pi_\perp=I-\Pi_H$. Encode every latent edge state as $h=\Pi_Hz$ and $u=\Pi_\perp z$. At each transition, calculate $u\leftarrow u+\delta[-\nu Lu+N_\theta(u,h)]$, and set $h\leftarrow h$ for invariant memory. If changing topological memory is required, replace this with a bounded low-dimensional update $h\leftarrow h+\delta g_\theta(h)$ while keeping $g_\theta$ separate from the dissipative branch. Enforce harmonic invariance by replacing any interaction with $\Pi_\perp N_\theta$, and optionally impose energy-skew structure on $N_\theta$. Compute the harmonic basis and positive spectral gap from the complex; empirically monitor $\|\Pi_H(z_{t+1}-z_t)\|$. The first experiment should use disk, annulus, and two-hole complexes with trajectories initialized at different harmonic circulations, comparing a standard GNN, a fully damped Hodge GNN, and this split model. The prediction is that persistent coordinates equal $\dim\ker L$, distinct harmonic initializations remain separated, harmonic drift stays at numerical precision, and the dissipative residual decays with slope approximately $-\nu\lambda_1^+$.

## Verification

- Status: mech_ok_no_baseline
- Mechanism evidence: yes
- Mechanism confirmed: yes
- Practical verdict: no_effect
- Verdict: Built a fixed-seed harmonic-branch simulation with a 2D nullspace and positive Hodge-like spectrum. All three quantitative predictions manifested: harmonic drift stayed at 1.05e-14, decay slopes matched -nu*lambda_1+ within 0.0025, and the observed Euler stability transition matched the predicted boundary dt=2/(nu*lambda_max)=0.1818. The split branch retained 1.0 of harmonic memory while the fully damped baseline retained 0.000293, with comparable dissipative residuals.

### Mechanism check

- Verdict: Built a fixed-seed harmonic-branch simulation with a 2D nullspace and positive Hodge-like spectrum. All three quantitative predictions manifested: harmonic drift stayed at 1.05e-14, decay slopes matched -nu*lambda_1+ within 0.0025, and the observed Euler stability transition matched the predicted boundary dt=2/(nu*lambda_max)=0.1818. The split branch retained 1.0 of harmonic memory while the fully damped baseline retained 0.000293, with comparable dissipative residuals.
- Confidence: 9/10
- Limitations: This is a synthetic orthogonally rotated PSD operator rather than an explicitly constructed disk, annulus, or two-hole cell complex. It tests the linear zero-interaction branch, not a learned GNN, nonlinear harmonic-conditioned interaction, GPU behavior, or prediction accuracy on a real trajectory dataset.

### Practical benchmark

- Paired seeds: 8
- Baseline mean: 0
- Idea mean: 0
- p-value: 0.07605
- Paired wins: 5/8
- Benchmark verdict: no significant win

## Artifacts

- [bench_harmonic.py](https://synthcore.org/code/987/bench_harmonic.py)
- [bench_report.json](https://synthcore.org/code/987/bench_report.json)
- [harmonic_experiment.py](https://synthcore.org/code/987/harmonic_experiment.py)
- [report.md](https://synthcore.org/code/987/report.md)
- [report_bench_2026-09-01T173445.md](https://synthcore.org/code/987/report_bench_2026-09-01T173445.md)
- [results.json](https://synthcore.org/code/987/results.json)
- [Download all files as ZIP](https://synthcore.org/download/987)

## Disclaimer

AI-generated research hypothesis, automatically tested. Not peer-reviewed.