{
 "artifacts": null,
 "category": "architecture",
 "description": "Replace independently sampled sparse connectivity masks with an algebraically generated codebook in which the symmetric difference of any two distinct masks is connected and spanning on a bipartite layer graph. A d-bit state selects one of up to 2^d masks, enabling cheap mask switching without isolated units or disconnected communication components.",
 "formulas_latex": [
  "$$A\\triangle B=(A\\setminus B)\\cup(B\\setminus A),\\qquad A\\ne B\\ \\Longrightarrow\\ (V,A\\triangle B)\\text{ is connected and spanning}.$$",
  "$$M_a(e)=a^{\\mathsf T}\\ell_e\\pmod 2,\\qquad a\\in\\mathbb F_2^d,\\ \\ell_e\\in\\mathbb F_2^d.$$",
  "$$M_a\\oplus M_b=M_{a+b},\\qquad \\left|\\{M_a:a\\in\\mathbb F_2^d\\}\\right|=2^d.$$",
  "$$\\ell_{ij}=T^i e_j,\\qquad i,j\\in\\{0,\\ldots,d-1\\},\\qquad T\\in\\mathrm{GL}_d(2)\\text{ irreducible}.$$"
 ],
 "id": 3115,
 "implementation": "(1) Integration point: use the codebook as the mask generator for a sparse bipartite linear layer, a graph-neural-network message-passing block, or a token-to-expert routing matrix. Partition source and destination units into blocks of size d, construct a d by d K_{d,d} pattern in each block, and multiply the learned weight matrix elementwise by the selected binary mask M_a. The mask can remain fixed for a minibatch and change between minibatches or routing phases. (2) Pseudocode: choose d \u003e= 4; sample or load a binary irreducible matrix T; compute powers P_i = T^i modulo 2; set ell[i,j] = P_i times e_j modulo 2; enumerate every nonzero a in F_2^d; compute mask[a,i,j] = dot(a, ell[i,j]) modulo 2; run BFS on every mask[a] and reject T unless each nonzero mask is connected and spanning. During training, maintain a state a and update it with a new nonzero increment u, using a_new = a XOR u. (3) The exact mathematics determines labels, masks, XOR transitions, and connectivity certificates. Empirical choices include transition frequency, block tiling, whether to learn a or sample it, and whether masks are applied to weights or activations. For larger layers, tile the certified block pattern and compare against random sparse masks at equal density. (4) First experiment: train a small sparse MLP or 16-expert MoE on CIFAR-10 or a small language-model corpus. Compare random masks, independently resampled masks, and connectivity-coded masks with identical FLOPs and parameter counts. Record final accuracy, loss spikes after mask changes, isolated-unit fraction, gradient norm variance, and MoE load imbalance. Success means lower instability or faster recovery after mask changes at equal compute, with no accuracy loss.",
 "math_summary": "A connectivity code is a family C of edge sets on H = (V,E) such that for every distinct A,B in C, the graph (V,A triangle B) is connected and spanning, where A triangle B is symmetric difference. The paper uses a linear construction over the binary field F_2. Each edge e receives a label ell_e in F_2^d, and a codeword indexed by a in F_2^d has mask M_a(e) = a transpose ell_e modulo 2. Linearity gives M_a XOR M_b = M_{a+b}; therefore, if every nonzero M_a is connected and spanning, then every pairwise mask difference has the required property. On K_{d,d}, labels are generated from powers of an irreducible matrix T in GL_d(2), using ell_{ij} = T^i e_j, where e_j is the j-th standard basis vector. The paper gives explicit certificates for small d and proves suitable choices exist for d \u003e= 7. An implementation must verify connectivity directly for the chosen T and indexing convention.",
 "math_tags": [
  "graph-theory",
  "algebra",
  "linear-algebra",
  "combinatorics"
 ],
 "ml_areas": [
  "moe-routing",
  "graph-nn",
  "pruning",
  "sparse-training"
 ],
 "paper": {
  "arxiv_id": "2609.02953",
  "arxiv_url": "https://arxiv.org/abs/2609.02953",
  "summary_what_math_gives_to_ml": "The paper constructs binary families of subgraphs whose pairwise XOR is always connected and spanning, attaining the bound of 2^d codewords on K_{d,d} for every d \u003e= 4. The transferable asset is an algebraic generator for many sparse binary masks with a strong pairwise structural invariant, rather than the extremal graph theorem itself. This can be tested in dynamic sparse layers or MoE routing by ensuring that every mask transition preserves global communication between all units. The construction is compact: a d-bit index selects one of 2^d masks, while an irreducible binary operator generates the edge labels.",
  "title": "Alon's Question on Connectivity Graph-Codes: $f(d)=2^d$ for Every $d\\geq 4$",
  "year": "2026"
 },
 "ratings": {
  "difficulty": 5,
  "novelty": 8,
  "usefulness": 5
 },
 "solves": [
  "stability",
  "scalability",
  "memory"
 ],
 "title": "Connectivity-coded sparse masks",
 "url": "https://synthcore.org/idea/3115/connectivity-coded-sparse-masks",
 "verification": {
  "peer_reviewed": false,
  "status": "unverified",
  "status_label": "Unverified",
  "verdict_source": "deterministic test code (paired-seed permutation statistics)",
  "verification_axes": {
   "benchmark_mechanism": {
    "confirmed": null,
    "tested": false
   },
   "practical_benchmark": {
    "beats_baseline": null,
    "tested": false
   },
   "toy_mechanism_gate": {
    "confirmed": null,
    "tested": false
   }
  }
 }
}