Matteo Sesia

Matteo Sesia

Assistant Professor at
DSO Department, USC Marshall

Publications

Training Uncertainty-Aware Classifiers with Conformalized Deep Learning

Published in pre-print, 2022

A flexible conformal inference method is developed to construct confidence intervals for the frequencies of queried objects in a very large data set, based on the information contained in a much smaller sketch of those data.

Recommended citation: Einbinder, Romano, Sesia, and Zhou (2022). "Training Uncertainty-Aware Classifiers with Conformalized Deep Learning." arXiv. https://arxiv.org/abs/2205.05878

Conformalized Frequency Estimation from Sketched Data

Published in pre-print, 2022

A flexible conformal inference method is developed to construct confidence intervals for the frequencies of queried objects in a very large data set, based on the information contained in a much smaller sketch of those data.

Recommended citation: Sesia and Favaro (2022). "Conformalized Frequency Estimation from Sketched Data." arXiv. https://arxiv.org/pdf/2204.04270

Hyperoxemia among Pediatric Intensive Care Unit Patients Receiving Oxygen Therapy

Published in Journal of Pediatric Intensive Care, 2021

This paper studies the prevalence and severity of hyperoxemia among pediatric intensive care unit patients at the Stanford Children’s Hospital.

Recommended citation: Fayazi, Azadeh R., Sesia, Matteo, Anand, Kanwaljeet J. S. (2021). "Hyperoxemia among Pediatric Intensive Care Unit Patients Receiving Oxygen Therapy." J Pediatr Intensive Care. https://doi.org/10.1055/s-0041-1740586

Conformal Prediction using Conditional Histograms

Published in Advances in Neural Information Processing Systems 33 (NeurIPS 2021, spotlight presentation), 2021

This paper develops efficient conformity scores for non-parametric regression.

Recommended citation: Sesia and Romano (2021). "Conformal Prediction using Conditional Histograms." Advances in Neural Information Processing Systems 33. https://arxiv.org/abs/2105.08747

Controlling the false discovery rate in GWAS with population structure

Published in Proc. Natl. Acad. Sci. U.S.A., 2020

A knockoff-based method for the genetic mapping of complex traits at multiple resolutions accounting for population structure, and a large-scale application to the UK Biobank data.

Recommended citation: Sesia, Bates, Candès, Marchini, and Sabatti (2020). "Controlling the false discovery rate in GWAS with population structure." Proc. Natl. Acad. Sci. U.S.A., 1180 (40). https://doi.org/10.1073/pnas.2105841118

A comparison of some conformal quantile regression methods

Published in Stat, 2020

This paper compares two recent methods that combine conformal inference with quantile regression to produce locally adaptive and marginally valid prediction intervals under sample exchangeability, both theoretically and empirically.

Recommended citation: Sesia and Candès. (2020). "A comparison of some conformal quantile regression methods." Stat. 9:e261. http://dx.doi.org/10.1002/sta4.261

Deep Knockoffs

Published in Journal of the American Statistical Association, 2019

This paper introduces a machine for sampling approximate model-X knockoffs for arbitrary and unspecified data distributions using deep generative models. By building upon the existing model-X framework, we thus obtain a flexible and model-free statistical tool to perform controlled variable selection.

Recommended citation: Romano, Sesia, and Candès. (2019). "Deep Knockoffs." J. Am. Stat. Assoc.. 0(0), pages 1-12. https://doi.org/10.1080/01621459.2019.1660174

Gene hunting with knockoffs for hidden Markov models

Published in Biometrika, 2019

This paper develop an exact construction of knockoffs for variables distributed as hidden Markov models, and builds upon this a principled and versatile method for controlling the false discovery rate in genome-wide association studies.

Recommended citation: Sesia, Sabatti, and Candès. (2019). "Gene hunting with knockoffs for hidden Markov models." Biometrika. 106(1), pages 1-18. https://doi.org/10.1093/biomet/asy033