pierre.bellec@rutgers.edu

Pierre C. Bellec

Assistant Professor (tenure-track), Department of Statistics, Rutgers University.

I am broadly interested in the properties of machine learning algorithms that extract structured information from noisy/corrupted data, with a focus on providing provable, certifiable guarantees on the output of these algorithms, e.g. with confidence intervals or other forms of uncertainty quantification.

Education

2016: PhD, ENSAE ParisTech, France, advised by Alexandre Tsybakov.
2012: Part III (MASt), University of Cambridge, UK.
2011: Diplôme d'Ingénieur, Ecole Polytechnique, France.

All publications

Preprints and submitted articles

Bibliography generated from build/preprints.bib
[1]
Derivatives and residual distribution of regularized M-estimators with application to adaptive tuning. Pierre C Bellec and Yiwei Shen.
arXiv:2107.05143, 2021.
[1]
Asymptotic normality of robust M-estimators with convex penalty. Pierre C Bellec, Yiwei Shen and Cun-Hui Zhang.
arXiv:2107.03826, 2021.
[1]
Out-of-sample error estimate for robust M-estimators with convex penalty. Pierre C Bellec.
arXiv:2008.11840, 2020.
[2]
Second order Poincare inequalities and de-biasing arbitrary convex regularizers when p/n → γ. Pierre C Bellec and Cun-Hui Zhang.
arXiv:1912.11943, 2019.
[3]
The noise barrier and the large signal bias of the Lasso and other convex estimators. Pierre C Bellec.
arXiv:1804.01230, 2018.
[5]
Optimistic lower bounds for convex regularized least-squares. Pierre C Bellec.
arXiv:1703.01332, 2017.
[6]
Concentration of quadratic forms under a Bernstein moment assumption. Pierre C. Bellec.
Technical report. Arxiv:1901.08726, 2014.

Journal articles

Bibliography generated from build/journals.bib
[4]
De-biasing the Lasso with degrees-of-freedom adjustment. Pierre C Bellec and Cun-Hui Zhang.
Bernoulli, to appear, 2021.
[1]
Second order Stein: SURE for SURE and other applications in high-dimensional inference. Pierre C Bellec and Cun-Hui Zhang.
Ann. Statist., to appear.
[2]
Optimal bounds for aggregation of affine estimators. Pierre C. Bellec.
Ann. Statist., 46(1):30–59, 02 2018.
[3]
Sharp oracle inequalities for Least Squares estimators in shape restricted regression. Pierre C. Bellec.
Ann. Statist., 46(2):745–780, 2018.
[4]
Slope meets Lasso: Improved oracle bounds and optimality. Pierre C. Bellec, Guillaume Lecué, and Alexandre B. Tsybakov.
Ann. Statist., 46(6B):3603–3642, 2018.
[5]
On the prediction loss of the lasso in the partially labeled setting. Pierre C. Bellec, Arnak S. Dalalyan, Edwin Grappin, and Quentin Paris.
Electron. J. Statist., 12(2):3443–3472, 2018.
[6]
Localized Gaussian width of M-convex hulls with applications to Lasso and convex aggregation. Pierre C Bellec.
Bernoulli, to appear, 2017.
[7]
Optimal exponential bounds for aggregation of density estimators. Pierre C. Bellec.
Bernoulli, 23(1):219–248, 2017.
[8]
Bounds on the prediction error of penalized least squares estimators with convex penalty. Pierre C Bellec and Alexandre B Tsybakov. In
Modern Problems of Stochastic Analysis and Statistics, Selected Contributions In Honor of Valentin Konakov. Springer, 2017.
[9]
Towards the study of least squares estimators with convex penalty. Pierre C Bellec, Guillaume Lecué, and Alexandre B Tsybakov. In
Seminaire et Congres, to appear, number 39. Societe mathematique de France, 2017.
[10]
A sharp oracle inequality for Graph-Slope. Pierre C. Bellec, Joseph Salmon, and Samuel Vaiter.
Electron. J. Statist., 11(2):4851–4870, 2017.
[11]
Adaptive confidence sets in shape restricted regression. Pierre C. Bellec.
Bernoulli, to appear, 2016.
[12]
Sharp Oracle Bounds for Monotone and Convex Regression Through Aggregation. Pierre C. Bellec and Alexandre B. Tsybakov.
Journal of Machine Learning Research, 16:1879–1892, 2015.

Conference proceedings

Bibliography generated from build/conferences.bib
[1]
Asymptotic normality and confidence intervals for derivatives of 2-layers neural network in the random features model. Pierre C Bellec and Yiwei Shen.
Accepted in Advances in Neural Information Processing Systems (NeurIPS), 2020, to appear.
[2]
The cost-free nature of optimally tuning Tikhonov regularizers and other ordered smoothers. Pierre C Bellec and Dana Yang.
Proceedings of the 37th International Conference on Machine Learning (ICML), pages 1621–1630, 2020.
[3]
First order expansion of convex regularized estimators. Pierre Bellec and Arun Kuchibhotla. In
Advances in Neural Information Processing Systems (NeurIPS), pages 3457–3468, 2019.
[4]
Aggregation of supports along the Lasso path. Pierre C. Bellec. volume 49 of
Proceedings of Machine Learning Research, pages 488–529, Conference On Learning Theory (COLT), Columbia University, New York, USA, 23–26 Jun 2016. PMLR.

Awards and Grants

  • NSF award DMS 1945428 (Principal Investigator): CAREER: Post-Differentiation Inference, 2020-2024.
  • NSF award DMS 1811976 (Principal Investigator): Uncertainty Quantification in High-Dimensional Structured Regression Problems, 2018-2021.
  • Blaise Pascal PhD Award, 2017 edition.

Student supervision

Courses and teaching material

Some of my teaching material is released under Creative Commons and available at https://github.com/bellecp/CC-BY-SA-teaching-material/ .

Past and upcoming talks

Professional Service

Software

Fast-p (https://github.com/bellecp/fast-p), a fast command-line tool to browse hundreds or thousands of academic PDFs.

Reach me

pierre.bellec@rutgers.edu     (Work)
Department of Statistics
Rutgers University
501 Hill Center, Busch Campus
110 Frelinghuysen Road
Piscataway, NJ 08854