465 Jon M. Huntsman Hall
3730 Walnut Street
Philadelphia, PA 19104
Research Interests: Statistics and machine learning
The two main interests in my group are:
The group is always looking to expand:
Seminar class in Fall 2019: Topics in Deep Learning (STAT-991), surveying advanced topics in deep learning research based on student presentations. See the Github page for the class materials.
Alnur Ali, Edgar Dobriban, Ryan J. Tibshirani The Implicit Regularization of Stochastic Gradient Flow for Least Squares.
Jonathan Lacotte, Sifan Liu, Edgar Dobriban, Mert Pilanci Limiting Spectrum of Randomized Hadamard Transform and Optimal Iterative Sketching Methods.
Sifan Liu and Edgar Dobriban (2019), Ridge Regression: Structure, Cross-Validation, and Sketching, International Conference on Learning Representations (ICLR) 2020.
Edgar Dobriban and Yue Sheng (Working), WONDER: Weighted one-shot distributed ridge regression in high dimensions.
Edgar Dobriban and Art B. Owen (2019), Deterministic parallel analysis: an improved method for selecting factors and principal components, Journal of the Royal Statistical Society, Series B , 81 (1), pp. 163-183.
Edgar Dobriban and Sifan Liu (2018), Asymptotics for Sketching in Least Squares Regression, Short version to appear at NeurIPS 2019.
Graphical displays; one- and two-sample confidence intervals; one- and two-sample hypothesis tests; one- and two-way ANOVA; simple and multiple linear least-squares regression; nonlinear regression; variable selection; logistic regression; categorical data analysis; goodness-of-fit tests. A methodology course. This course does not have business applications but has significant overlap with STAT 101 and 102.
This page has links to methods from my papers. Feel free to contact me if you are interested to use them.
The ePCA method for principal component analysis of exponential family data, e.g. Poisson-modeled count data. (with L.T. Liu);
Methods for working with large random data matrices, including
P-value weighting techniques for multiple hypothesis testing. These can improve power in multiple testing, if there is prior information about the individual effect sizes. Includes the iGWAS method for Genome-Wide Association Studies.