Noureddine El Karoui
high-dimensional statistics, random matrices, high-dimensional robust regression, high-dimensional M-estimation, the bootstrap and resampling in high-dimension, limit theorems and statistical inference, applied statistics, auction theory from the bidder standpoint, fairness, market design, control, reinforcement learning, statistics/ML/AI at industrial scale, design of experiments, game theory
In recent years I have worked as a Principal Researcher at Criteo AI Lab and a Principal Staff Software Engineer at LinkedIn.
My research interests include/have included high-dimensional statistics, random matrices, high-dimensional robust regression, high-dimensional M-estimation, the bootstrap and resampling in high-dimension, limit theorems and statistical inference, as well as applied statistics and connections to Finance. I am currently also working on auction theory (especially from the bidder standpoint), fairness in Machine Learning, Market Design, control, learning to rank, and reinforcement learning. I am especially interested in industrial-scale statistics/machine learning/AI, opportunities and limitations therein, and connections to Economics.
I am a recipient of a Sloan Research Fellowship (Mathematics), an IMS Fellowship, and was an invited speaker at the 2018 International Congress of Mathematicians (ICM; probability and statistics section). My research has been supported by NSF grants (including a CAREER grant) and a Sloan research fellowship. I am grateful for their support.
Brief bio:
I did my undergraduate studies at Ecole Polytechnique, in France, majoring in Applied Mathematics. I then studied at Stanford, where I got a PhD in Statistics (co-advised by David Donoho and Iain Johnstone) and a Master's in Financial Mathematics. After my Phd, I did a 1-year post-doc in genetic epidemiology (advised by Alice Whitemore) and then joined UC Berkeley.
- Tracy-Widom limit for the largest eigenvalue of a large class of complex sample covariance matrices,The Annals of Probabililty,35(2): 663--714, March 07
- A rate of convergence result for the largest eigenvalue of complex white Wishart matrices, The Annals of Probability, 34(6):2077--2117, November 06
- Recent results about the largest eigenvalue of random covariance matrices and statistical application, Acta Physica Polonica B, 36(9):2681-2697, September 2005
- Getting more from digital SNP data (With Wei Zhou and Alice Whittemore),Statistics in Medicine 25:3124-3133, September 2006
- On the largest eigenvalue of Wishart matrices when n,p and p/n tend to infinity, Unpublished, Available on arxiv.org
- Spectrum estimation for large dimensional covariance matrices using random matrix theory, Annals of Statistics, 36(6): 2757-2790, December 2008
- Operator norm consistent estimation of large dimensional sparse covariance matrices, Annals of Statistics, 36(6): 2717-2756, December 2008
- Concentration of measure and spectra of random matrices: applications to correlation matrices, elliptical distributions and beyond, Annals of Applied Probability, 19(6):2362-2405, December 2009
- The spectrum of kernel random matrices, Annals of Statistics, 38(1): 1-51, February 2010
- High-dimensionality effects in the Markowitz problem and other quadratic programs with linear constraints: risk underestimation, Annals of Statistics, 38(10):3487–3566, December 2010
- On information plus noise kernel random matrices, Annals of Statistics, 38(10):3191–3216, October 2010
- Chapter « Random matrix Theory », Encyclopedia of Quantitative Finance, Publisher : Wiley, Editor : Rama Cont
- Chapter « Multivariate Statistics », Handbook of Random Matrix Theory, Publisher: Oxford; Editors: G. Akemann, J. Baik,, P. Di Francesco
- On the realized risk of Markowitz portfolios, SIAM Journal on Financial Mathematics
- Second order accurate distributed eigenvector computation for extremely large matrices (with Alexandre d’Aspremont), Electronic Journal of Statistics, 4(2010), 1345-1385
- Geometric sensitivity of random matrix results: consequences for shrinkage estimators of covariance and related statistical methods, (with Holger Koesters); under revision (67 pages)
- Weak recovery conditions from graph portioning bounds and order statistics (with Alexandre d’Aspremont), Mathematics of Operations Research, 38,(2); 228-247, May 2013
- On robust regression with high-dimensional predictors (with Bean, Bickel,Lim and Yu), PNAS, 2013 110 (36) (August, 2013) 14557-14562
- Penalized robust regression in high-dimension (with Bean, Bickel, Lim, Yu). Tech report 813 (2011)
- Optimal M-estimation in high-dimensional regression (with Bean, Bickel and Yu), PNAS, 2013 110 (36) (August, 2013) 14563-14568
- Optimizing Automated Classification of Periodic Variable Stars in New Synoptic Surveys (with Long (1st author), Rice, Richards, Bloom), Publications of the Astronomical Society of the Pacific, 124 (913); March 2012, 280-295
- Estimation error reduction in portfolio optimization with Conditional Value-at-Risk (with Andrew Lim and Gah-Yi Vahn), 2nd round of revision, Management Science (33 pages)
- A stochastic smoothing algorithm for semi-definite programming (with Alexandre d’Aspremont), SIAM Journal in Optimization 2014, 24 (3), pp. 1138-117
- Asymptotic behavior of unregularized and ridge-regularized high-dimensional robust regression estimators : rigorous results; Arxiv: 1311.2445
- Random matrices and high-dimensional M-estimation: applications to robust regression, penalized robust regression and GLMs (Video; Harvard Applied Math Colloquium, March 2014)
- Vector diffusion maps and random matrices with random blocks (with Hau-tieng Wu); Information and Inference (arXiv:1310.0188)
- Kernel density estimation with Berkson error (with Long(1st author) and Rice); Submitted (arXiv:1401.3362)
- Graph Connection Laplacian methods can be made robust to noise (with Hau-tieng Wu); To appear in Annals of Statistics (arXiv:1405.6231)
- Can we trust the bootstrap in high-dimension? The case of linear models (with Elizabeth Purdom); JMLR
- On the impact of predictor geometry on the performance of high-dimensional ridge-regularized generalized robust regression estimators; PTRF
- The bootstrap, covariance matrices, and PCA in moderate and high-dimensions (with Elizabeth Purdom); Submitted
- Explicit shading strategies in repeated truthful auctions (with Abeille, Calauzenes, Nedelec, Perchet)
- Thresholding the virtual value: a simple method to increase welfare and lower reserve prices in online auction systems (with Nedelec, Abeille, Calauzenes, Heymann, Perchet)
- Asymptotics For High Dimensional Regression M-Estimates: Fixed Design Results. (with Lei and Bickel); PTRF
- Random matrices and high-dimensional statistics: beyond covariance matrices. Proceedings of the International Congress of Mathematicians, Rio, 2018