The paper introduces μs’ autotune, an automatic tuning strategy for the Lasso that optimizes a penalized Gaussian log‑likelihood over regression coefficients and noise standard deviation. Extensive simulations on regression and VAR models show that autotune is faster and yields better generalization and model selection, especially in low signal‑to‑noise regimes. The method also delivers a new noise‑standard‑deviation estimator, a visual diagnostic for sparsity, and is demonstrated on a real‑world financial dataset, with an accompanying R package available on GitHub.
By Tathagata Sadhukhan, Ines Wilms, Stephan Smeekes, Sumanta Basu
The paper introduces a feature selection framework for high‑dimensional financial regression that uses Monte Carlo Tree Search with Generalized Nested Rollout Policy Adaptation. It treats feature selection as a sequential decision process and evaluates candidate subsets by out‑of‑sample validation performance, incorporating subset size and multicollinearity into a lexicographic objective. Applied to a NASDAQ‑100 weekly return regression with 1,072 engineered features, the method outperforms an optimized Lasso baseline, achieving an out‑of‑sample R² of 0.6961 versus 0.6313.
By Cameron Mouangue, Tristan Cazenave
arXiv:2309. 15769v3 Announce Type: replace-cross Abstract: Recent advances in deep learning have highlighted the phenomenon of benign overfitting in overparameterized statistical models, sparking significant interest in understanding its foundations.
By Dennis Shen, Dogyoon Song, Peng Ding, Jasjeet S. Sekhon
The paper investigates regression when outcomes are observed only after passing through selection filters that depend on both covariates and outcomes, a common issue in fields such as clinical trials, labor markets, and auctions. It provides a complete characterization of the minimal assumptions on the functional forms of selection processes that allow regression to remain possible, and shows that the regression function can sometimes be identified even when the selection filter itself cannot. Under stronger identification conditions, the authors also deliver finite‑sample estimation guarantees, explicit convergence rates, and oracle‑efficient algorithms, offering the first general‑purpose estimation method for this broad class of selection problems.
By Vikram Kher, Jane H. Lee, Anay Mehrotra, Manolis Zampetakis
arXiv:2608.22597v1 Announce Type: new
Abstract: Subsampling is effective in tackling computational challenges for massive data with rare events. Overly aggressive subsampling may adversely affect est...
By Jing Wang, HaiYing Wang, Qiang Zhang, Hao Helen Zhang
arXiv:2306. 14851v5 Announce Type: replace-cross Abstract: Given a high-dimensional covariate matrix and a response vector, ridge-regularized sparse linear regression selects a subset of features that explains the relationship between covariates and the response in an interpretable manner.
By Ryan Cory-Wright, Andr\'es G\'omez
The paper proposes a unified framework for estimating sparse precision and covariance matrices by reparametrizing the last column of either matrix. This reparametrization reduces the likelihood optimization to a one‑dimensional minimization followed by a penalized multiple regression, applicable to any penalty separable in off‑diagonal entries and linear on the diagonal. When the penalty is the α1 norm, the method simultaneously recovers the graphical Lasso and the covariance graphical Lasso algorithms.
By Aramayis Dallakyan, Mohsen Pourahmadi
arXiv:2405. 03063v3 Announce Type: replace-cross Abstract: We propose a generalized debiased Lasso estimator based on a stability principle.
By Jingbo Liu
The paper addresses the problem of covariate mismatch in collaborative linear prediction, where different centers collect different sets of features. It proposes estimators that leverage information across centers under center-wise MCAR missingness, including a plug‑in estimator for low dimensions and an impute‑then‑regress strategy with ridge regularization for higher dimensions. All methods comply with federated learning constraints, exchanging only aggregated statistics, and the authors provide theoretical learning rates and numerical validation.
By Alexis Ayme, R\'emi Khellaf
The paper investigates feature priming in high‑dimensional online linear regression, showing that estimating feature weights from past data and refitting a minimum‑norm predictor can lead to regret that scales with sparsity rather than ambient dimension. It provides a negative answer to a COLT 2023 open problem by proving that three natural priming rules incur ≥Ω(min{T,√d}) regret against a zero‑loss one‑sparse comparator, due to cheap nuisance interpolation that underweights truly predictive coordinates. The authors also identify conditions under which regret is governed by data rank and present constructions that achieve tight univariate rates, while noting that the multivariate case remains unresolved.
By Huibo Xu, Shi Fu, Qixin Zhang, Dacheng Tao
arXiv:1907.06994v2 Announce Type: replace-cross
Abstract: Mixtures of experts (MoE) are conditional mixture models in which both the mixing proportions and the component densities depend on the predi...
By Thin Nguyen-Van, Faicel Chamroukhi, Ha Hoang Van, Bao Tuyen Huynh
arXiv:2610.07637v1 Announce Type: cross
Abstract: Many real-world datasets exhibit unusually large values far more frequently than predicted by Gaussian models. Heavy-tailed distributions capture thi...
By Kaito Takanami, Takashi Takahashi, Yoshiyuki Kabashima