SearcharxivSearch

arXiv subjects

Sudhakar Gadde

Publications and source records attributed to Sudhakar Gadde.

2 recordsLinked to original sources

Morita Transport, Marked Projective Loci, and Dedekind Classification for C4-Type Conditionss

Objectwise Morita invariance of $\Cfour$ is known. We prove the corresponding statements for $\Cfourstar$, semi-weak-CS, and strongly $\Cfourstar$, including the chain joins in the definition of semi-weak-CS. For a module property $\mathcal Q$ preserved and reflected by equivalences, we consider the subset $\Vmon_{\mathcal Q}(R)$ of the finitely generated projective monoid $\Vmon(R)$. This marked pair is transported by Morita equivalence. The associated regular-module ring property is Morita invariant if and only if membership in $\Vmon_{\mathcal Q}(R)$ is constant on the order units. The matrix profile satisfies $\mu_{\mathcal Q}(M_m(R))=\{n\geq1:mn\in\mu_{\mathcal Q}(R)\}$. For a nondivision right Ore domain, the profiles of $\Cfour$, $\Cfourstar$, and strongly $\Cfourstar$ are $\{1\}$, whereas the semi-weak-CS profile is $\mathbb N_{\geq1}$. Over a nonfield Dedekind domain $D$, the first three marked loci consist of the projectives of rank at most one, while every finitely generated projective is semi-weak-CS. For a finitely generated $D$-module $M$, we prove that every $\Cfour$ defect at rank at most one can be transferred between $M$ and $\tau(M)$, and we construct a defect when the rank is at least two. Hence $M$ is $\Cfour$ if and only if $\operatorname{rank}_D M\leq1$ and $\tau(M)$ is $\Cfour$; the analogous criterion holds for $\Cfourstar$. Combined with the finite-length torsion criterion, these reductions give invariant-factor tests for the four conditions and criteria for their behaviour under direct sums.

math.RA

Certified Learning under Distribution Shift: Sound Verification and Identifiable Structure

Proposition. Let $f$ be a predictor trained on a distribution $P$ and evaluated on a shifted distribution $Q$. Under verifiable regularity and complexity constraints, the excess risk under shift admits an explicit upper bound determined by a computable shift metric and model parameters. We develop a unified framework in which (i) risk under distribution shift is certified by explicit inequalities, (ii) verification of learned models is sound for nontrivial sizes, and (iii) interpretability is enforced through identifiability conditions rather than post hoc explanations. All claims are stated with explicit assumptions. Failure modes are isolated. Non-certifiable regimes are characterized.

cs.LG