One-extra-dimension conjecture for complete-graph squared-stress optimization
For every number of points , integers with , and configurations and , define
If and the Hessian is positive semidefinite, then . Equivalently, every second-order stationary point satisfies for all ; this holds without genericity assumptions and allows repeated points and degenerate configurations.
References
Primary source
Additional references
- One Extra Dimension Suffices for the Complete-Graph Squared-Stress — arXiv — Lilin Yan, Hongwei Zhao
Progress summary
A recent paper proves the conjecture after doubling the required dimension, but the one-extra-dimension case remains open.
The conjecture asserts that complete-graph squared-stress optimization has no nonglobal second-order critical points once the ambient dimension is one larger than the configuration dimension. The exact threshold remains unresolved.
Known results
- The landscape can have spurious local minima when , even with points (Song et al., 2025; Criscitiello et al., 2026).
- Benignness is proved at the conjectured threshold when .
- For arbitrary numbers of points, every second-order critical point is globally optimal when (Criscitiello, 2026).
October 2026 formalization claim
A separate arXiv item by Lilin Yan and Hongwei Zhao advertises a Lean 4 formalization, kernel check, and axiom audit for a claimed theorem concerning this setting. The retrieved material does not independently assess whether the encoded statement matches the conjecture or whether it establishes the full claim.
Current status (as of October 2026): The complete-graph conjecture for remains open; the factor-two theorem and the restricted case are established, while the formalization-based full-resolution claim is unverified.
Solutions 0
No solutions have been posted yet.