# Boundary-Only Cell-Complex Network

- ID: 719
- Canonical URL: https://synthcore.org/idea/719/boundary-only-cell-complex-network
- API JSON: https://synthcore.org/api/idea/719.json
- API Markdown: https://synthcore.org/api/idea/719.md
- Verification status: unverified
- Source: [arXiv:2607.10950](https://arxiv.org/abs/2607.10950)
- Category: architecture
- Solves: memory, stability, generalization
- ML areas: graph-nn, embedding, memory, regularization
- Math tags: algebraic-topology, linear-algebra, graph-theory, algebra
- Ratings: usefulness 6/10; difficulty 5/10; novelty 7/10

## Idea description

Parameterize a cell-complex neural network by features on p-cells and derive lower-dimensional boundary features using the cellular boundary map over F2. For a 2D square complex, neighboring plaquette bits determine each link feature through XOR, reproducing the paper's exact gauge-law reconstruction and preventing the network from representing inconsistent open boundary configurations.

## Mathematical statement

The paper solves the one-form Gauss law and obtains the operator identity tau^x_l = sigma^x_{p_L} sigma^x_{p_R}, where l is a link and p_L,p_R are its two incident plaquettes; multiplication is the Z2 group operation. Equivalently, represent plaquette states by a binary vector f in F2^{n_2}, let B_2 in F2^{n_1 x n_2} be the cellular boundary/incidence matrix from plaquettes to links, and reconstruct link states as b = B_2 f mod 2. For a square lattice, b_l = f_{p_L} XOR f_{p_R}. The chain-complex identity B_1 B_2 = 0, where B_1 maps links to vertices, guarantees that the reconstructed link configuration has zero vertex boundary and therefore consists of closed domain-wall loops. In a differentiable network, replace hard bits by signed features s in [-1,1] and use the relaxed product s_l = s_{p_L}s_{p_R}; the exact binary boundary map is used at sampling or inference time.

## Key formulas

- $$\tau^{x}_{\ell}=\sigma^{x}_{p_{L}}\sigma^{x}_{p_{R}}$$
- $$b=B_{2}f\pmod 2,\qquad b_{\ell}=f_{p_L}\oplus f_{p_R}$$
- $$B_{1}B_{2}=0$$
- $$s_{\ell}=s_{p_L}s_{p_R},\qquad s_p\in[-1,1]$$

## Implementation notes

1. Integration point: use this as the input parameterization and message-passing interface of a graph neural network on an oriented 2D cell complex. Store encoder-produced plaquette features H2 in R^{n2 x d}; do not independently store link features. Construct the sparse plaquette-to-link incidence matrix B2, with one row per link and entries identifying its incident plaquettes. For binary Z2 data, compute XOR exactly. For continuous features, use signed plaquette states S2 = tanh(H2) and reconstruct each link feature by elementwise multiplication of its incident plaquette states; concatenate this with fixed link attributes before link processing. 2. Pseudocode: `H2 = encoder(plaquette_input); S2 = tanh(H2); Sl[e] = product(S2[p] for p in incident_plaquettes[e]); H1 = link_mlp(concat(Sl, link_attributes)); H2_next = plaquette_update(H2, aggregate_incident(H1)); output = readout(H2,H1)`. For discrete sampling, use `f = Bernoulli(sigmoid(H2))` and `b = (B2 @ f) % 2`. 3. Compute B2 and B1 directly from the mesh; no learning or estimation is required. Verify the hard constraint by measuring `max(abs((B1 @ b) % 2))`, which must be zero. Compare against a baseline that independently predicts link and plaquette features and adds a Gauss-law penalty. 4. First experiment: train a small cell-complex GNN on synthetic 16x16-grid and triangulated-mesh tasks such as boundary-loop prediction from plaquette fillings. Report parameter count, activation memory, test accuracy, mesh-size transfer, and exact constraint violations. The method succeeds if it has zero violations without penalty, lower memory at equal hidden width, and equal or better accuracy on unseen mesh topologies. A useful stress test is random plaquette masking followed by boundary-loop reconstruction.

## Disclaimer

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