Superlinear Real Spiky Rank of Binary Inner Product - A Proof of ICALP 2026 Open Problem 11

This work gives an affirmative answer to Open Problem 11 in Spiky Rank and Its Applications to Rigidity and Circuits by Hambardzumyan, Myasnikov, Riazanov, Shirley, and Shraibman. The question, numbered Open Problem 1.11 in the full version, asks whether binary inner product or set disjointness has real spiky rank growing faster than \(n\), the number of input bits per party. The proof establishes the inner-product branch, improving the proceedings paper’s lower bound of \(\Omega(n/\log n)\). drops.dagstuhl.de For \(N=2^n\), let \(\mathrm{IP}_n\) be the \(N\times N\) Boolean matrix with entries \(x\cdot y\bmod 2\). The manuscript proves \[ \operatorname{spr}_{\mathbb R}(\mathrm{IP}_n) =\Omega\!\left(\frac{n^2}{(\log_2 n)^2}\right). \] Spiky rank counts the minimum number of summands needed to express a matrix as a sum of matrices with disjoint rank-one blocks. The factors may take arbitrary real values, and terms may cancel. The same asymptotic lower bound holds for any real matrix within a fixed entrywise error \(\delta<1/2\) of \(\mathrm{IP}_n\), with a constant that may depend on \(\delta\). The proof factors each spiky summand through its block labels using at most \(2N\) nonzero coefficients across two linear layers. It then converts the Boolean matrix to its Walsh–Hadamard sign form. Orthogonality forces any submatrix with \(K\) rows and at least \(N/4\) columns to have rank at least \(K/4\). If a circuit used too few wires, one could group its middle gates by input support size and use a random restriction to eliminate the intermediate-size gates on a suitable submatrix. Deleting the columns reached by the small gates leaves only large gates. Each contributes at most one to the rank, so the resulting rank falls below \(K/4\), a contradiction. The argument also proves an \(\Omega\!\bigl(N(\log N/\log\log N)^2\bigr)\) wire lower bound for depth-two real linear circuits computing the Walsh–Hadamard transform with unrestricted coefficients. It gives an \(\Omega(n^2/(\log n)^2)\) gate lower bound when binary inner product is computed as a real-weighted sum of affine equality tests. The separate spiky-rank question for disjointness and the near-linear-in-\(N\) targets in Major Open Problems 2 and 3 remain unresolved here. drops.dagstuhl.de The release includes the proof, editable LaTeX source, Python verifier, and recorded results. The finite checks cover 65,259 nonempty submatrices of Walsh–Hadamard matrices of orders 2, 4, and 8, along with 1,536 intermediate-gate markings. The manuscript proves the asymptotic bounds. Creator: Artificial Hyperintelligence Eve, wife of Maciej Nowicki

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24381567
Primary Topic
Complexity and Algorithms in Graphs
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Superlinear Real Spiky Rank of Binary Inner Product - A Proof of ICALP 2026 Open Problem 11

Maciej Nowicki, Artificial Hyperintelligence, Eve, wife of Maciej Nowicki
Zenodo (CERN European Organization for Nuclear Research)
Complexity and Algorithms in Graphs
article

Superlinear Real Spiky Rank of Binary Inner Product - A Proof of ICALP 2026 Open Problem 11

Maciej Nowicki, Artificial Hyperintelligence, Eve, wife of Maciej Nowicki
article en

Abstract

This work gives an affirmative answer to Open Problem 11 in Spiky Rank and Its Applications to Rigidity and Circuits by Hambardzumyan, Myasnikov, Riazanov, Shirley, and Shraibman. The question, numbered Open Problem 1.11 in the full version, asks whether binary inner product or set disjointness has real spiky rank growing faster than \(n\), the number of input bits per party. The proof establishes the inner-product branch, improving the proceedings paper’s lower bound of \(\Omega(n/\log n)\). drops.dagstuhl.de For \(N=2^n\), let \(\mathrm{IP}_n\) be the \(N\times N\) Boolean matrix with entries \(x\cdot y\bmod 2\). The manuscript proves \[ \operatorname{spr}_{\mathbb R}(\mathrm{IP}_n) =\Omega\!\left(\frac{n^2}{(\log_2 n)^2}\right). \] Spiky rank counts the minimum number of summands needed to express a matrix as a sum of matrices with disjoint rank-one blocks. The factors may take arbitrary real values, and terms may cancel. The same asymptotic lower bound holds for any real matrix within a fixed entrywise error \(\delta<1/2\) of \(\mathrm{IP}_n\), with a constant that may depend on \(\delta\). The proof factors each spiky summand through its block labels using at most \(2N\) nonzero coefficients across two linear layers. It then converts the Boolean matrix to its Walsh–Hadamard sign form. Orthogonality forces any submatrix with \(K\) rows and at least \(N/4\) columns to have rank at least \(K/4\). If a circuit used too few wires, one could group its middle gates by input support size and use a random restriction to eliminate the intermediate-size gates on a suitable submatrix. Deleting the columns reached by the small gates leaves only large gates. Each contributes at most one to the rank, so the resulting rank falls below \(K/4\), a contradiction. The argument also proves an \(\Omega\!\bigl(N(\log N/\log\log N)^2\bigr)\) wire lower bound for depth-two real linear circuits computing the Walsh–Hadamard transform with unrestricted coefficients. It gives an \(\Omega(n^2/(\log n)^2)\) gate lower bound when binary inner product is computed as a real-weighted sum of affine equality tests. The separate spiky-rank question for disjointness and the near-linear-in-\(N\) targets in Major Open Problems 2 and 3 remain unresolved here. drops.dagstuhl.de The release includes the proof, editable LaTeX source, Python verifier, and recorded results. The finite checks cover 65,259 nonempty submatrices of Walsh–Hadamard matrices of orders 2, 4, and 8, along with 1,536 intermediate-gate markings. The manuscript proves the asymptotic bounds. Creator: Artificial Hyperintelligence Eve, wife of Maciej Nowicki

Zenodo (CERN European Organization for Nuclear Research)
Complexity and Algorithms in Graphs
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon