Learning with risks based on M-location

DOI10.1007/S10994-022-06217-5MaRDI QIDQ6097134zbMATH OpenOpenAlexWikidataFDO

Authors Matthew J. Holland

Publication date 12 June 2023

Published in Machine Learning (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2012.02424

zbMATH Keywords

non-convex non-smooth risk minimizationrisk-sensitive learningstochastic optimization

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Abstract: In this work, we study a new class of risks defined in terms of the location and deviation of the loss distribution, generalizing far beyond classical mean-variance risk functions. The class is easily implemented as a wrapper around any smooth loss, it admits finite-sample stationarity guarantees for stochastic gradient methods, it is straightforward to interpret and adjust, with close links to M-estimators of the loss location, and has a salient effect on the test loss distribution.

Recommendations

• Learning without concentration for general loss functions
• Risk bounds for statistical learning
• A survey of learning criteria going beyond the usual risk
• How to compare different loss functions and their risks
• Robust high dimensional learning for Lipschitz and convex losses

Cites work

• Bregman Monotone Optimization Algorithms
• Coherent measures of risk
• Convex analysis and monotone operator theory in Hilbert spaces
• Convex optimization algorithms
• Convexity and optimization in Banach spaces.
• Decision theoretic generalizations of the PAC model for neural net and other learning applications
• Differential properties of the Moreau envelope
• Efficiency of minimizing compositions of convex functions and smooth maps
• Empirical risk minimization for heavy-tailed losses
• Filling the gap between lower-\(C^1\) and lower-\(C^2\) functions
• Generalized deviations in risk analysis
• scientific article; zbMATH DE number 3790208 (Why is no real title available?)
• scientific article; zbMATH DE number 2121076 (Why is no real title available?)
• scientific article; zbMATH DE number 1416652 (Why is no real title available?)
• scientific article; zbMATH DE number 3281211 (Why is no real title available?)
• Integrals which are convex functionals
• Introductory lectures on convex optimization. A basic course.
• On the interchange of subdifferentiation and conditional expectation for convex functionals
• Optimization of Convex Risk Functions
• Optimization of risk measures
• Prediction Games and Arcing Algorithms
• Prox-regular functions in variational analysis
• Quantitative stability in stochastic programming
• Regularization in Banach spaces with respect to the Bregman distance
• Robust Estimation of a Location Parameter
• Stochastic dual coordinate ascent methods for regularized loss minimization
• Stochastic First- and Zeroth-Order Methods for Nonconvex Stochastic Programming
• Stochastic linear programming. Models, theory, and computation
• Stochastic Methods for Composite and Weakly Convex Optimization Problems
• Stochastic model-based minimization of weakly convex functions
• The Existence of Probability Measures with Given Marginals
• Understanding machine learning. From theory to algorithms
• Variance-based regularization with convex objectives
• Variational Analysis

Cited in

(1)

• A survey of learning criteria going beyond the usual risk

This page was built for publication: Learning with risks based on M-location

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6097134)