TY - GEN
T1 - HappyMap
T2 - 14th Innovations in Theoretical Computer Science Conference, ITCS 2023
AU - Deng, Zhun
AU - Dwork, Cynthia
AU - Zhang, Linjun
N1 - Publisher Copyright:
© Zhun Deng, Cynthia Dwork, and Linjun Zhang; licensed under Creative Commons License CC-BY 4.0.
PY - 2023/1/1
Y1 - 2023/1/1
N2 - Multicalibration is a powerful and evolving concept originating in the field of algorithmic fairness. For a predictor f that estimates the outcome y given covariates x, and for a function class C, multi-calibration requires that the predictor f(x) and outcome y are indistinguishable under the class of auditors in C. Fairness is captured by incorporating demographic subgroups into the class of functions C. Recent work has shown that, by enriching the class C to incorporate appropriate propensity re-weighting functions, multi-calibration also yields target-independent learning, wherein a model trained on a source domain performs well on unseen, future, target domains (approximately) captured by the re-weightings. Formally, multicalibration with respect to C bounds E(x,y)∼D[c(f(x), x) · (f(x) − y)] for all c ∈ C. In this work, we view the term (f(x) − y) as just one specific mapping, and explore the power of an enriched class of mappings. We propose s-Happy Multicalibration, a generalization of multi-calibration, which yields a wide range of new applications, including a new fairness notion for uncertainty quantification, a novel technique for conformal prediction under covariate shift, and a different approach to analyzing missing data, while also yielding a unified understanding of several existing seemingly disparate algorithmic fairness notions and target-independent learning approaches. We give a single HappyMap meta-algorithm that captures all these results, together with a sufficiency condition for its success.
AB - Multicalibration is a powerful and evolving concept originating in the field of algorithmic fairness. For a predictor f that estimates the outcome y given covariates x, and for a function class C, multi-calibration requires that the predictor f(x) and outcome y are indistinguishable under the class of auditors in C. Fairness is captured by incorporating demographic subgroups into the class of functions C. Recent work has shown that, by enriching the class C to incorporate appropriate propensity re-weighting functions, multi-calibration also yields target-independent learning, wherein a model trained on a source domain performs well on unseen, future, target domains (approximately) captured by the re-weightings. Formally, multicalibration with respect to C bounds E(x,y)∼D[c(f(x), x) · (f(x) − y)] for all c ∈ C. In this work, we view the term (f(x) − y) as just one specific mapping, and explore the power of an enriched class of mappings. We propose s-Happy Multicalibration, a generalization of multi-calibration, which yields a wide range of new applications, including a new fairness notion for uncertainty quantification, a novel technique for conformal prediction under covariate shift, and a different approach to analyzing missing data, while also yielding a unified understanding of several existing seemingly disparate algorithmic fairness notions and target-independent learning approaches. We give a single HappyMap meta-algorithm that captures all these results, together with a sufficiency condition for its success.
KW - algorithmic fairness
KW - target-independent learning
KW - transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85147538387&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85147538387&partnerID=8YFLogxK
U2 - 10.4230/LIPIcs.ITCS.2023.41
DO - 10.4230/LIPIcs.ITCS.2023.41
M3 - Conference contribution
AN - SCOPUS:85147538387
T3 - Leibniz International Proceedings in Informatics, LIPIcs
BT - 14th Innovations in Theoretical Computer Science Conference, ITCS 2023
A2 - Kalai, Yael Tauman
PB - Schloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing
Y2 - 10 January 2023 through 13 January 2023
ER -