# Magnitude-Euler Path Signature Regularizer

- ID: 1317
- Canonical URL: https://synthcore.org/idea/1317/magnitude-euler-path-signature-regularizer
- API JSON: https://synthcore.org/api/idea/1317.json
- API Markdown: https://synthcore.org/api/idea/1317.md
- Verification status: unverified
- Source: [arXiv:2607.23357](https://arxiv.org/abs/2607.23357)
- Category: regularization
- Solves: accuracy, generalization, stability
- ML areas: mlp, regularization, training-dynamics, graph-nn
- Math tags: metric-geometry, topology, combinatorics, graph-theory
- Ratings: usefulness 5/10; difficulty 6/10; novelty 8/10

## Idea description

Represent the computation graph of an MLP as a directed acyclic Lawvere metric space and compute a truncated, length-resolved Euler signature of its active paths. Add a penalty that separates signatures between classes while suppressing signatures that are insensitive to labels, thereby encouraging globally distinct computation routes without changing layer widths or degree statistics.

## Mathematical statement

The paper defines an extended quasi-pseudometric d:X×X→[0,∞], where asymmetry and d(x,x)≠0 are allowed, and a k-simplex x^(k)=(x_0,…,x_k) with adjacent points distinct and length λ(x^(k))=Σ_{j=1}^k d(x_{j-1},x_j). The magnitude chain group at length L is MC_{k,L}(X)=R{x^(k): λ(x^(k))=L}, with differential ∂_k=Σ_{j=1}^{k-1}(-1)^j∂^(j). The jth face is retained only when the middle point is metrically redundant: ∂^(j)(x^(k))=∇_j x^(k) if d(x_{j-1},x_{j+1})=d(x_{j-1},x_j)+d(x_j,x_{j+1}), and is zero otherwise. For a finite graph, the truncated length-L Euler characteristic is χ_L(X)=Σ_{k=0}^{K}(-1)^k rank(MC_{k,L}(X)); this alternating chain count is the decategorified quantity used here. For implementation, X is the set of neurons and input/output terminals, d(u,v) is the shortest directed path cost, R=R, and only lengths L≤L_max and simplex orders k≤K are retained. With unit edge cost, L is an integer hop length; with positive edge costs c_uv, d(u,v)=min_{p:u↝v}Σ_{(a,b)∈p}c_ab.

## Key formulas

- $$\lambda(x^{(k)})=\sum_{j=1}^{k}d(x_{j-1},x_j),\qquad MC_{k,L}(X):=R\{x^{(k)}:\lambda(x^{(k)})=L\}.$$
- $$\partial_k=\sum_{j=1}^{k-1}(-1)^j\partial^{(j)},\qquad \partial^{(j)}(x^{(k)})=\begin{cases}\nabla_jx^{(k)}&d(x_{j-1},x_{j+1})=d(x_{j-1},x_j)+d(x_j,x_{j+1}),\\0&\text{otherwise.}\end{cases}$$
- $$\chi_L(X)=\sum_{k=0}^{K}(-1)^k\,\operatorname{rank}(MC_{k,L}(X)),\qquad d(u,v)=\min_{p:u\leadsto v}\sum_{(a,b)\in p}c_{ab}.$$
- $$\mathcal L_{\mathrm{ME}}=\mathcal L_{\mathrm{task}}+\beta\sum_{L=0}^{L_{\max}}\left[\operatorname{Var}_{y}\bigl(\chi_L(X\mid y)\bigr)+\rho\,\bigl(\chi_L(X)-\bar\chi_L\bigr)^2\right].$$

## Implementation notes

1. Integration point: apply this during training to a small auxiliary computation graph extracted from each MLP block. Create one node for every neuron in the selected hidden layers, plus source and class-output nodes. For a dense layer from neuron u to v, use an edge cost c_uv=1 for the first experiment; optionally use c_uv=1/(|w_uv|+epsilon) or c_uv=softplus(-|w_uv|) after the unit-cost baseline. Remove edges whose gated activation contribution |w_uv h_u| is below a percentile threshold, producing a sample-dependent active DAG. 2. Pseudocode: build active edges; compute shortest directed distances d[u,v] by dynamic programming in topological order; for each L≤L_max, enumerate or sample ordered tuples (x_0,…,x_k) with adjacent finite distances and sum d[x_{j-1},x_j]=L; accumulate n[k,L]. Set chi[L]=sum_k (-1)^k n[k,L], directly implementing the paper’s Euler decategorification. Add the loss above, where chi(X|y) is the minibatch mean for class y, bar_chi is an exponential moving-average target, beta is small, and rho optionally stabilizes the global signature. 3. Mathematics versus estimates: shortest distances and chain lengths are exact for the chosen graph; chi is estimated by uniform chain sampling when enumeration exceeds a budget. Use K=3, L_max=4, 256 sampled chains per k and class. Since hard thresholding is nondifferentiable, update this statistic every N steps and treat it as a detached monitoring or reinforcement-style regularizer initially; a differentiable surrogate can replace edge inclusion with sigmoid gates and soft chain weights. 4. First experiment: train a 3-layer width-256 MLP on MNIST or CIFAR-10 embeddings, comparing cross-entropy, weight decay, and the Euler-signature penalty at matched FLOPs. Report test accuracy, calibration, classwise signature separation, and robustness to neuron permutation or random pruning. A positive result is higher accuracy or corruption robustness at equal parameter count, with classwise chi vectors separating before accuracy improves and no more than 10% training-time overhead.

## Disclaimer

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