arXiv subjects

Suvrit Sra

Publications and source records attributed to Suvrit Sra.

At least 19 recordsLinked to original sources

arXiv2026

On some positivity properties of the Riemann $Ξ$ function

We prove strict log-concavity of the positive functions $Θ$ and $Λ$ with one-sided and bilateral Laplace transforms $1/ξ(\tfrac12+\sqrt s)$ and $1/Ξ$, respectively, where $Ξ(z)=ξ(\tfrac12+iz)$. For either function, log-concavity is the Pólya frequency property of order two (PF$_2$). Without assuming the Riemann hypothesis (RH), but using its published verification up to height about $3\cdot10^{12}$, we further prove that both densities are PF$_{k_*}$ for $k_*=1\,810\,540\,931\,075$. The proof views $Θ$ as a small signed perturbation of a density that is PF of every order, with an error bound independent of the order, so the numerical check reduces to scalar inequalities. Since $Λ$ is a Gaussian mixture of $Θ$, it inherits the same order. Conversely, a pole obstruction shows that if either density is PF$_k$, then every zeta zero lower than the $(k+2)$-nd lies on the critical line. Finally, we prove that if RH holds up to height $T>\tfrac12$, then $Ξ$ satisfies the generalized Laguerre inequalities $L_n[Ξ](x)\ge0$ of every order $n$ for $|x|\le T-\tfrac12$, extending the case $n=1$ of Csordas, Ruttan and Varga (1991).

math.CA↗

arXiv2026

Linearly Convergent Algorithms for Nonsmooth Problems with Unknown Smooth Regions

We develop efficient algorithms for optimizing piecewise smooth (PWS) functions when their smooth regions are \emph{unknown} to the algorithm. For convex PWS functions satisfying a quadratic growth (QG) condition, we demonstrate that the bundle-level (BL)-type method~\cite{lemarechal1995new} achieves global linear convergence---to our knowledge, the first such result for any algorithm for this problem class. We extend the method to solve weakly convex PWS problems, achieving the same order of complexity as in smooth non-convex optimization. Furthermore, we introduce a verifiable and accurate termination criterion for PWS optimization. Like the gradient norm in smooth optimization, this certificate tightly characterizes the optimality gap under the QG condition and can be evaluated without knowledge of any problem parameters. We develop a search subroutine for this certificate and embed it within a guess-and-check framework, resulting in almost parameter-free algorithms for both convex QG and weakly convex PWS problems.

math.OC↗

arXiv2026

The $p$-Bures-Wasserstein distance via $3\times3$ block matrices (with application to $p$-Bures-Wasserstein barycenters)

We study the $p$-Bures-Wasserstein distance $D_p(A,B)=(\mathrm{tr}\,A^p+\mathrm{tr}\,B^p-2\|A^{1/2}B^{1/2}\|_p^p)^{1/2}$ on positive semidefinite matrices; it equals the Bures-Wasserstein distance for $p=1$, and the Hilbert-Schmidt distance for $p=2$. We prove that $D_p$ is a metric for every $1\le p\le2$, by proving a stronger 2014 conjecture of Lin and van den Driessche on $3\times3$ block matrices. Our main tool is a partition inequality for Schatten norms, proved by complex interpolation. We also characterize Fréchet means for $D_p$ by a fixed-point equation and prove geometric convergence of the associated iteration.

math.FA↗

arXiv2026

A Proof of McLeod's 1959 Conjecture on Complete Homogeneous Symmetric Ratios

We prove McLeod's (1959) conjecture that $(h_{\ell+r}/h_\ell)^{1/r}$ is subadditive for all $\ell\ge0$ and $r\ge1$, where $h_k$ is the complete homogeneous symmetric polynomial of degree $k$ (J. B. McLeod, "On four inequalities in symmetric functions", Proc. Edinburgh Math. Soc., 1959). We first prove that the ratio $h_k/h_{k-1}$ is convex, and our proof takes an unexpected detour through queuing theory. The geometric mean of $r$ consecutive such ratios equals McLeod's conjectured function, but geometric means of convex functions need not be convex. Our second and decisive step is a tensorization trick that couples neighboring ratios tightly enough to make their geometric mean convex.

math.CA↗

arXiv2026

A Koteljanskii inequality for permanents

We prove a permanental analogue of Koteljanskii's inequality. If $A$ is an inverse $M$-matrix that becomes symmetric after a positive diagonal similarity, then $\mathrm{per}(A_{S\cup T})\,\mathrm{per}(A_{S\cap T})\ge\mathrm{per}(A_S)\,\mathrm{per}(A_T)$ for all $S,T\subseteq[n]$, where $A_S$ is the principal submatrix indexed by $S$. The proof expresses permanents as moments of a complex Gaussian vector and uses Ginibre's correlation inequality. Finally, an explicit counterexample shows that symmetry cannot be dropped.

math.RA↗

arXiv2026

Positive definite functions of noncommuting contractions, Hua-Bellman matrices, and a new distance metric

We study positive definiteness of Hua-Bellman matrices (i.e., matrices of the form $[\det(I-A_i^*A_j)^{-α}]_{ij}$, where $A_i$ and $A_j$ are strict contractions and $α\in\mathbb{R}$). We start by revisiting a 1959 work of Bellman (Representation theorems and inequalities for Hermitian matrices; Duke Mathematical J., 26(3), 1959) that studies positive definiteness of Hua-Bellman matrices and claims a strengthening of Hua's representation-theoretic results (Inequalities involving determinants; Acta Mathematica Sinica, 5 (1955)). We uncover a critical error in Bellman's proof that has surprisingly escaped notice to date, and we show by an explicit example that his claim itself is false. We then provide conditions under which $\det(I-A^*B)^{-α}$ is a positive definite function. For complex contractions our condition is the known sharp one, with the integer exponents below Hua's range reached by an elementary argument. For real symmetric contractions we recover the part of Bellman's claims that survives. Building on our result, we introduce a new Poincaré-like distance metric on noncommuting strict contractions, and yet another `log-det' based distance in the appendix.

math.FA↗

arXiv2026

Tighter bounds on Komlós discrepancy: existence and algorithmic results

Guo, Fang, and Lu recently proved the Komlós conjecture: for vectors $v_1,\ldots,v_n\in\mathbb R^d$ of Euclidean norm at most one, there are signs $\varepsilon_j\in\{-1,1\}$ with $\|\sum_j\varepsilon_jv_j\|_\infty\le3\sqrt{2π}$. We give a short proof of the bound $3π$ that keeps the geometric lifting framework of (Guo, Fang, and Lu 2026a) while replacing the analytic core by a quadratic Dirichlet energy. Moreover, using a more fine-grained analysis of the stability of the product-cosine function of (Smirnov and Vershynin 2026) under translations, we further sharpen the bound to below $6.9013$. On the algorithmic side, based on the polynomial-time construction of (Guo, Fang, and Lu 2026b) we give a deterministic algorithm that finds a coloring of discrepancy at most $37.54$ using at most $\widetilde O(mn+n^4)$ arithmetic operations.

math.CO↗

arXiv2026

Percolation Dynamics in Optimization: Variance Cascades and Nested Symmetry

We study the dynamics of Stochastic Gradient Descent (SGD), which is known to steer deep neural networks toward invariant sets that correspond to simpler subnetworks. How this steering unfolds over time remains poorly understood. We answer this by modeling the stochastic gradient flow (SGF) as a percolation process, in which nested architectural symmetries force subnetworks to merge in discrete blocks rather than by single-edge attachment. These structural transitions register as variance spikes in a macroscopic order parameter echoing physical phase transitions. We further state sufficient conditions under which the trapping argument carries over to Adam and AdamW under heavy-tailed gradient noise and measure them on a trained Transformer.

cs.LG↗

arXiv2026

Kadison--Singer partitions and Bilu--Linial graph signings in polynomial time

We prove two main algorithmic results in spectral discrepancy. First, we give a deterministic polynomial-time rounding theorem for rational positive semidefinite matrices of arbitrary rank. The algorithm starts from any rational fractional signing and assigns one sign per original matrix. Its discrepancy is less than $3.37\,\|\sum_i \mathrm{Tr}(A_i)A_i\|^{1/2}$. This yields Kadison--Singer half-partitions with error below $1.69\sqrt{\varepsilon}$, as well as deterministic graph signings that control signed adjacency and signed degrees simultaneously. The proof builds on the spectral-potential method of Ezeunala and Jiang (2026) and introduces a new way to choose rounding directions. We prove polynomial bit complexity for the rounding procedure. Second, we give a Las Vegas algorithm for the Bilu--Linial signing problem on an arbitrary prescribed graph. If $G$ has $n$ vertices and maximum degree $Δ\ge3$, the algorithm terminates almost surely. It uses fewer than $100n^3$ insertion attempts in expectation and returns a signing with $\|A_s\|<2\sqrt{2(Δ-1)}$. For bipartite graphs its one-sided form gives the sharp universal bound $\|A_s\|<2\sqrt{Δ-1}$. The algorithm builds the signing by inserting vertices and recursively deleting and restoring neighbors after rejected insertions. In the analysis, the $\sqrt2$ gap to the Bilu--Linial conjecture comes from a factor of two in the bound for vertex deletions in the two-sided case. On a $d$-regular bipartite Ramanujan base the same signing produces a Ramanujan $2$-lift of that prescribed base.

cs.DS↗

arXiv2026

Boolean Small-Ball Inequalities for Discrepancy Theory

We prove new small-ball inequalities for boolean matrix-series. The leading example is $\mathbb E_s[{\text{det}(I-S^2)^β\,\mathbf 1_{\{\|S\|<1\}}}]\ge e^{-O(βτ)}$, which holds for boolean matrix-series $S=\sum_i s_iA_i$ formed using symmetric matrices $A_1,\dots,A_n$ and uniformly random signs $s\in\{\pm1\}^n$. Specifically, this inequality holds for all $β\ge1$ with $τ=\sum_i\text{Tr} A_i^2$, as soon as the maximum of $(\text{Tr} A_i^2)_{i=1}^n$ and a certain variance term are bounded above by universal constants. The proof combines the Gaussian reciprocal estimate of (Akbas and Sra 2026), the directional-variation signing theorem of (Guo, Fang, and Lu 2026), and a replica argument that turns existence into a Gibbs law on good signings. Most notably, boolean small-ball delivers a new, interlacing-free proof of Kadison-Singer (most general case); it also recovers Matrix Spencer and Komlós as quick corollaries, while yielding more than six almost immediate proofs of an assortment of discrepancy theoretic problems.

math.PR↗

arXiv2026

GPT, the Counterexample Machine

This document reports over 15 counterexamples found using GPT Pro over the course of 12 months, for problems in combinatorics, number theory, convexity, analysis, and other areas. The counterexamples concern both (reasonably) well-known problems as well as lesser known instances. The counterexamples are accompanied by the repo https://github.com/suvrit/count-ex-machina, and we warmly welcome new contributions toward helping building a rich \emph{``open-source countex library.''}

math.CO↗

arXiv2026

A Proof of the Matrix Spencer Conjecture

We develop a novel approach to matrix discrepancy based on matrix small-ball estimates. Specifically, we use a determinantal weight (obtained from the log-barrier) to scale the small-ball probability into a partition function of a tilt of the Gaussian measure. We then employ matrix-weighted Poincaré inequalities to compare this partition function to that of a pinched or diagonal part of the matrix, obtaining \emph{dimension-free} constants. Our technique yields a hereditary small-ball estimate for Gaussian series that should be of independent interest. As the main application, we resolve the Matrix Spencer conjecture: for symmetric $n\times n$ matrices $A_1,\dots,A_n$ with $\|A_i\|\le1$, one can efficiently find a coloring $x\in\{\pm1\}^n$ with $\|\sum_{i=1}^n x_iA_i\|=O(\sqrt n)$.

math.FA↗

arXiv2026

An Algebraic Matrix Spencer Theorem

We develop an algebraic approach to matrix discrepancy based on the representation theory of finite-dimensional C$^*$-algebras. As an application, we resolve a substantial structured special case of the Matrix Spencer conjecture. In particular, we show that for every family of contractions $A_1,\ldots,A_n$ that are contained in a finite-dimensional $C^*$-algebra $\mathcal A$ with $\text{dim}_{\mathbb C} (\mathcal A) \lesssim n$, there exists signs $x\in\{\pm1\}^n$ such that $\|\sum_{i=1}^n x_i A_i\| \le O(\sqrt n)$. As a noteworthy special case, our main result also resolves the Group Spencer conjecture of (Bandeira'24). We furthermore prove that Matrix Spencer continues to hold for low-rank perturbations of matrix families coming from an $C^*$-algebra of small dimension.

math.PR↗

arXiv2026

Free Heavy-Tailed Lunch for Muon: A Theoretical Justification of Empirical Success

Non-Euclidean optimisation methods with matrix-valued updates, such as Muon and Scion, have recently shown strong empirical performance for training Transformer models, yet their theoretical advantages over Euclidean methods remain poorly understood. We address this gap in the heavy-tailed non-convex regime, where stochastic gradients have bounded $p$-th central moments, $p \in (1,2]$. We show that certain non-Euclidean methods achieve optimal sample complexity under stronger stationarity measures, while Euclidean methods incur additional dimension-dependent costs. As a consequence, for $m \times n$ matrices, Muon finds an $\varepsilon$-stationary point in nuclear norm within $\mathcal{O}\left(\min\{m, n\} \frac{Δ_1 L}{\varepsilon^2} \left(\frac σ\varepsilon \right)^{\frac p {p-1}}\right)$ samples, absorbing heavy-tailed noise without extra dimension dependence, unlike Euclidean methods. We further prove this sample complexity, including its dimension dependence, is optimal for all first-order methods under nuclear-norm stationarity. Experiments on large language models support our theory. Surprisingly, our results suggest that other Schatten geometries beyond the spectral geometry of Muon can perform competitively in certain settings.

math.OC↗

arXiv2026

Active Inference as Context Acquisition for AI Agents

Interactive AI agents must acquire the right context as efficiently as possible. When a user omits a constraint, preference, file, or task variable, an agent can proceed with a default assumption or spend tokens on a clarifying question, retrieval call, tool call, or prompt trial. We formulate this tradeoff as active inference for context acquisition. An inner inference step updates beliefs over a latent task state, and an outer decision selects the next context action, task action, or stop action to minimize expected free energy under cost. In deterministic settings, the epistemic term reduces to expected information gain, optionally normalized by token cost. We instantiate the framework in Optimal Question Asking (OQA), with exact posteriors and a dynamic programming oracle, and benchmark frontier language models on binary and multiway categorical tasks from 25 to 300 candidates. We also study clarification before generation and automated prompt optimization under token budgets. The formulation is model-agnostic and views active inference as a design principle for the context-acquisition layer of AI agents.

cs.AI↗

arXiv2026

Trees to Flows and Back: Unifying Decision Trees and Diffusion Models

Decision trees and diffusion models are ostensibly disparate model classes, one discrete and hierarchical, the other continuous and dynamic. This work unifies the two by establishing a crisp mathematical correspondence between hierarchical decision trees and diffusion processes in appropriate limiting regimes. Our unification reveals a shared optimization principle: \emph{Global Trajectory Score Matching (GTSM)}, for which gradient boosting (in an idealized version) is asymptotically optimal. We underscore the conceptual value of our work through two key practical instantiations: \treeflow, which achieves competitive generation quality on tabular data with higher fidelity and a 2\times computational speedup, and \dsmtree, a novel distillation method that transfers hierarchical decision logic into neural networks, matching teacher performance within 2\% on many benchmarks.

cs.LG↗

arXiv2026

A projection-based framework for gradient-free and parallel learning

We present a feasibility-seeking approach to neural network training. This mathematical optimization framework is distinct from conventional gradient-based loss minimization and uses projection operators and iterative projection algorithms. We reformulate training as a large-scale feasibility problem: finding network parameters and states that satisfy local constraints derived from its elementary operations. Training then involves projecting onto these constraints, a local operation that can be parallelized across the network. We introduce PJAX, a JAX-based software framework that enables this paradigm. PJAX composes projection operators for elementary operations, automatically deriving the solution operators for the feasibility problems (akin to autodiff for derivatives). It inherently supports GPU/TPU acceleration, provides a familiar NumPy-like API, and is extensible. We train diverse architectures (MLPs, CNNs, RNNs) on standard benchmarks using PJAX, demonstrating its functionality and generality. Our results show that this approach is a compelling alternative to gradient-based training, with clear advantages in parallelism and the ability to handle non-differentiable operations.

cs.LG↗

arXiv2026

Tight Generalization Bounds for Noiseless Inverse Optimization

Inverse optimization (IO) seeks to infer the parameters of a decision-maker's objective from observed context--action data. We study noiseless IO, where demonstrations are generated by a ground-truth objective. We provide a high-probability ${O}(\frac{d}{T})$ generalization bound for the induced action set, where $d$ is the number of unknown parameters and $T$ is the size of the training dataset. We strengthen these guarantees under additional conditions that ensure uniqueness of the chosen action, bringing our IO guarantees in line with best-arm identification results in the bandit literature. We further show that the ${O}(\frac{d}{T})$ rate is tight over all consistent estimators considered here, and extend the result to both instantaneous and cumulative regret. Notably, the resulting regret lower bound matches the corresponding upper bounds in the adversarial setting, indicating that the stochastic IO setting is effectively adversarial for the class of estimators studied here. Finally, we propose a parameter-free algorithm with lower per-iteration complexity than generic solvers. Experiments validate the predicted rates and illustrate the tightness of our bounds.

stat.ML↗