Fast and Robust Least Squares Estimation in Corrupted Linear Models

Abstract

Subsampling methods have been recently proposed to speed up least squares estimation in large scale settings. However, these algorithms are typically not robust to outliers or corruptions in the observed covariates. The concept of influence that was developed for regression diagnostics can be used to detect such corrupted observations as shown in this paper. This property of influence -- for which we also develop a randomized approximation -- motivates our proposed subsampling algorithm for large scale corrupted linear regression which limits the influence of data points since highly influential points contribute most to the residual error. Under a general model of corrupted observations, we show theoretically and empirically on a variety of simulated and real datasets that our algorithm improves over the current state-of-the-art approximation schemes for ordinary least squares.

References in corpus (5)

• High-dimensional regression with noisy and missing data: Provable guarantees with nonconvexity
• Randomized algorithms for matrices and data
• A Statistical Perspective on Algorithmic Leveraging
• Orthogonal Matching Pursuit with Noisy and Missing Data: Low and High Dimensional Results
• Improved matrix algorithms via the Subsampled Randomized Hadamard Transform

Cited by in corpus (7)

• Robust Regression via Hard Thresholding
• Byzantine Fault Tolerant Distributed Linear Regression
• More Efficient Estimation for Logistic Regression with Optimal Subsample
• Robust Gaussian Process Regression for Real-Time High Precision GPS Signal Enhancement
• Inverse Reinforcement Learning from a Gradient-based Learner
• One Backward from Ten Forward, Subsampling for Large-Scale Deep Learning
• Adaptive Iterative Hessian Sketch via A-Optimal Subsampling