Distributed dual averaging algorithm's convergence rate scales inversely with network spectral gap, enabling efficient large-scale optimization.

John Duchi, Alekh Agarwal, Martin Wainwright

Homophily and contagion are generically confounded in observational social network studies, requiring strong assumptions to distinguish.

Cosma Rohilla Shalizi, Andrew C. Thomas

Optimization under unknown constraints using Gaussian processes and Bayesian learning, applied to healthcare policy.

Robert B. Gramacy, Herbert K. H. Lee

Randomized truthful mechanism via single call to monotone allocation rule.

Moshe Babaioff, Robert D. Kleinberg, Aleksandrs Slivkins

Introduces LinUCB, a linear contextual Bandit algorithm, achieving 12.5% click rate lift on Yahoo! dataset, suitable for large-scale personalized news recommendation.

Lihong Li, Wei Chu, John Langford et al.

Constructs adaptive confidence bands in density estimation using wavelet/kernel methods, leveraging extreme value theory to ensure honest coverage over a 'generic' subset.

Evarist Giné, Richard Nickl

Exp4.P algorithm achieves supervised learning-like guarantees in contextual bandit problems, significantly reducing regret.

Alina Beygelzimer, John Langford, Lihong Li et al.

This paper links test martingales with Bayes factors and p-values, introducing functions to limit evidence exaggeration, enabling systematic conversion between them.

Glenn Shafer, Alexander Shen, Nikolai Vereshchagin et al.

Proposes a universal framework using local explanation vectors to analyze classification decisions for any classifier.

David Baehrens, Timon Schroeter, Stefan Harmeling et al.

Randomized algorithms for low-rank matrix approximation enable fast, robust processing of massive data sets, outperforming classical methods in speed and scalability.

Nathan Halko, Per-Gunnar Martinsson, Joel A. Tropp

Using CoRoT data, identified 358 candidate B pulsators, revealing a new class of low-amplitude B-type pulsators.

P. Degroote, C. Aerts, M. Ollivier et al.

Proposed quantum algorithm prepares thermal Gibbs state with time upper bound D^α, α linked to free energy density.

David Poulin, Pawel Wocjan

Proposed a fast QMA amplification method using quantum reflection and walk, doubling speed over Marriott and Watrous' method.

Daniel Nagaj, Pawel Wocjan, Yong Zhang

Robust matrix completion via nuclear norm minimization, achieves accurate recovery from nr log^2 n noisy samples with error proportional to noise level.

Emmanuel J. Candes, Yaniv Plan

Proposes nonparanormal model using smooth transformations for high-dimensional sparse graph estimation, outperforming Gaussian models on non-Gaussian data.

Han Liu, John Lafferty, Larry Wasserman

The study uses stellar evolution code to simulate stellar collisions in young clusters, finding no intermediate-mass black holes.

E. Glebbeek, E. Gaburov, S. E. de Mink et al.

Proposes a spectral matrix completion algorithm achieving O(rn) sample efficiency with provable error bounds.

Raghunandan H. Keshavan, Andrea Montanari, Sewoong Oh

Unified f-divergences, Bregman divergences, and derived new SVM formulation via integral and variational representations.

Mark D. Reid, Robert C. Williamson

Linearly Parameterized Bandits use exploration-exploitation strategy to achieve Θ(r√T) cumulative regret and Bayes risk.

Paat Rusmevichientong, John N. Tsitsiklis

Introduces block coherence and guarantees exact recovery of block-sparse signals via BOMP and ℓ2/ℓ1 optimization under specific conditions.

Yonina C. Eldar, Helmut Bolcskei