FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3199071121> ?p ?o ?g. }

• W3199071121 abstract "Exponential tilting is a technique commonly used in fields such as statistics, probability, information theory, and optimization to create parametric distribution shifts. Despite its prevalence in related fields, tilting has not seen widespread use in machine learning. In this work, we aim to bridge this gap by exploring the use of tilting in risk minimization. We study a simple extension to ERM -- tilted empirical risk minimization (TERM) -- which uses exponential tilting to flexibly tune the impact of individual losses. The resulting framework has several useful properties: We show that TERM can increase or decrease the influence of outliers, respectively, to enable fairness or robustness; has variance-reduction properties that can benefit generalization; and can be viewed as a smooth approximation to the tail probability of losses. Our work makes rigorous connections between TERM and related objectives, such as Value-at-Risk, Conditional Value-at-Risk, and distributionally robust optimization (DRO). We develop batch and stochastic first-order optimization methods for solving TERM, provide convergence guarantees for the solvers, and show that the framework can be efficiently solved relative to common alternatives. Finally, we demonstrate that TERM can be used for a multitude of applications in machine learning, such as enforcing fairness between subgroups, mitigating the effect of outliers, and handling class imbalance. Despite the straightforward modification TERM makes to traditional ERM objectives, we find that the framework can consistently outperform ERM and deliver competitive performance with state-of-the-art, problem-specific approaches." @default.
• W3199071121 created "2021-09-27" @default.
• W3199071121 creator A5008645615 @default.
• W3199071121 creator A5046195860 @default.
• W3199071121 creator A5068602599 @default.
• W3199071121 creator A5082543438 @default.
• W3199071121 date "2021-09-13" @default.
• W3199071121 modified "2023-09-23" @default.
• W3199071121 title "On Tilted Losses in Machine Learning: Theory and Applications" @default.
• W3199071121 cites W1494085563 @default.
• W3199071121 cites W1541527977 @default.
• W3199071121 cites W1616087023 @default.
• W3199071121 cites W1647779468 @default.
• W3199071121 cites W1786332878 @default.
• W3199071121 cites W1964191893 @default.
• W3199071121 cites W1970136394 @default.
• W3199071121 cites W1975736346 @default.
• W3199071121 cites W1977691343 @default.
• W3199071121 cites W1987793616 @default.
• W3199071121 cites W1989684337 @default.
• W3199071121 cites W2008740515 @default.
• W3199071121 cites W2019291268 @default.
• W3199071121 cites W2023320048 @default.
• W3199071121 cites W2031355801 @default.
• W3199071121 cites W2040949867 @default.
• W3199071121 cites W2046033161 @default.
• W3199071121 cites W2054321814 @default.
• W3199071121 cites W2059120410 @default.
• W3199071121 cites W2069627663 @default.
• W3199071121 cites W2085988980 @default.
• W3199071121 cites W2088413745 @default.
• W3199071121 cites W2103012681 @default.
• W3199071121 cites W2108263314 @default.
• W3199071121 cites W2112796928 @default.
• W3199071121 cites W2120340025 @default.
• W3199071121 cites W2123097583 @default.
• W3199071121 cites W2123838014 @default.
• W3199071121 cites W2125607229 @default.
• W3199071121 cites W2132984949 @default.
• W3199071121 cites W2133472795 @default.
• W3199071121 cites W2135094946 @default.
• W3199071121 cites W2139914196 @default.
• W3199071121 cites W2142901448 @default.
• W3199071121 cites W2151976295 @default.
• W3199071121 cites W2157017541 @default.
• W3199071121 cites W2165203736 @default.
• W3199071121 cites W2183341477 @default.
• W3199071121 cites W2205115547 @default.
• W3199071121 cites W2530395818 @default.
• W3199071121 cites W2540757487 @default.
• W3199071121 cites W2541884796 @default.
• W3199071121 cites W2562677074 @default.
• W3199071121 cites W2577784528 @default.
• W3199071121 cites W2604763608 @default.
• W3199071121 cites W2621265919 @default.
• W3199071121 cites W2751645773 @default.
• W3199071121 cites W2755098789 @default.
• W3199071121 cites W2765393021 @default.
• W3199071121 cites W2808674837 @default.
• W3199071121 cites W2943927551 @default.
• W3199071121 cites W2944407464 @default.
• W3199071121 cites W2946307404 @default.
• W3199071121 cites W2947761153 @default.
• W3199071121 cites W2953963175 @default.
• W3199071121 cites W2962688627 @default.
• W3199071121 cites W2962935454 @default.
• W3199071121 cites W2963081269 @default.
• W3199071121 cites W2963092340 @default.
• W3199071121 cites W2963248893 @default.
• W3199071121 cites W2963275611 @default.
• W3199071121 cites W2963322627 @default.
• W3199071121 cites W2963351448 @default.
• W3199071121 cites W2963371670 @default.
• W3199071121 cites W2963476860 @default.
• W3199071121 cites W2963516811 @default.
• W3199071121 cites W2963539647 @default.
• W3199071121 cites W2963759070 @default.
• W3199071121 cites W2963803533 @default.
• W3199071121 cites W2963818033 @default.
• W3199071121 cites W2963999980 @default.
• W3199071121 cites W2964097078 @default.
• W3199071121 cites W2964106499 @default.
• W3199071121 cites W2964135521 @default.
• W3199071121 cites W2964241437 @default.
• W3199071121 cites W2964273174 @default.
• W3199071121 cites W2964309657 @default.
• W3199071121 cites W2970416783 @default.
• W3199071121 cites W2970711760 @default.
• W3199071121 cites W2970762077 @default.
• W3199071121 cites W2971239241 @default.
• W3199071121 cites W2978290191 @default.
• W3199071121 cites W2985955583 @default.
• W3199071121 cites W2994873335 @default.
• W3199071121 cites W2995315671 @default.
• W3199071121 cites W2995735216 @default.
• W3199071121 cites W2996860127 @default.
• W3199071121 cites W3007729382 @default.
• W3199071121 cites W3011766119 @default.
• W3199071121 cites W3014704188 @default.
• W3199071121 cites W3034986954 @default.

• next