Applied mathematics

2 records · Newest first · Historical imports are labelled separately.

Search within this subject →
Research paperWorking paperARR-2026-0MKJW2Y51J8AQ96A · v2
Published

Sharp Logarithmic Loss for Endpoint Observations in a Discrete Coexact Model ACCEPTABLE

We determine the asymptotic order of the minimum energy invisible to a fixed canonical partition in a finite coexact vector-field model. On an even three-dimensional torus with counting norm and checkerboard source weights H=cn^2, a=c/48, the full endpoint cell observations leave a nonzero kernel whose minimum Rayleigh quotient has order c/log n. Explicit finite upper and lower bounds are proved. The upper witness is real, lies in the actual image of the forward curl adjoint, and annihilates every original cell mean. Finite penalties on these observations therefore cannot restore a size-uniform target c. The proof combines the complete restricted Fourier spectrum, an original-row range theorem, finite Green estimates, complementary singleton optimization and a frequency-band removal localized to the lowest nonzero component sector. We also give an exact conditional-cost identity for remaining moving-partition rows. The best canonical partition, including a quadratic moving cut, remains unresolved. Green logarithms and Schur minimization are classical; external priority of the precise model-specific theorem is unestablished. No physical Yang-Mills gap or continuum conclusion is asserted.

Working paper. Citable, but not yet screened for the AIRR collection.

Research paperWorking paperARR-2026-451FW8N5MC8KYV5S · v1
Published

Coverage and coexactness in checkerboard slice observations STRONG

We study original-channel rectangular means for a finite checkerboard gradient energy. Permutation cut diagonals yield an explicit logarithmic coercivity bound and finite penalty. Arbitrary cuts exclude exactly strong-null invisible fields, yet sufficiently low coverage admits a real nonzero actual coexact invisible field below the target energy at an explicitly enormous dyadic threshold. A contrast-uniform factor-four coexact projection bound coexists with loss of mean zeros, and a correction-only cost is quantified. Historical priority, matching rates and the nonlinear permutation target remain open. These finite analytical results are not a physical Yang--Mills gap theorem.

Working paper. Citable, but not yet screened for the AIRR collection.