Paulo Orenstein

PhD candidate in Statistics

Stanford University

I am a fifth-year PhD student in the Department of Statistics at Stanford University, advised by Persi Diaconis. My research broadly focuses on the interplay between statistics, probability, and computation, particularly as they apply to high-dimensional Bayesian models and Monte Carlo methods. I enjoy working on both theoretical and applied projects, and find them to often illuminate each other.

I spent the 2018 summer working as a research intern at Oculus Research, in the eye-tracking team. Before coming to Stanford, I received a Masters in Mathematics and BSc in Economics, both from PUC-Rio, in Brazil.

Interests

High-dimensional Bayesian Statistics
Monte Carlo Methods
Probability Theory
Applied Statistics and Machine Learning

Education

PhD Candidate in Statistics

Stanford University
Masters in Mathematics, 2014

PUC-Rio
BSc in Economics, 2012

PUC-Rio

Papers

S2S Reboot: an Argument for Greater Inclusion of Machine Learning in Subseasonal to Seasonal Forecasts

with Judah Cohen, Dim Coumou, Jessica Hwang, Lester Mackey, Sonja Totz, and Eli Tziperman
In WIREs Climate Change, 2018

Finite-sample Guarantees for Winsorized Importance Sampling

Preprint Code Slides

Bayes Shrinkage at GWAS Scale: Convergence and Approximation Theory of a Scalable MCMC Algorithm for the Horseshoe Prior

with James Johndrow and Anirban Bhattacharya
Preprint, 2018

Preprint Slides

Improving Subseasonal Forecasting in the Western U.S. with Machine Learning

with Jessica Hwang, Karl Pfeiffer, Judah Cohen and Lester Mackey
Preprint, 2018

Preprint Code Dataset Poster Slides News (USBR) News (Microsoft)

When (and How) to Favor Incumbents in Optimal Dynamic Procurement Auctions

with Vinicius Carrasco and Pablo Salgado
In Journal of Mathematical Economics, 2016

Talks

Winsorized Importance Sampling

Theory and results for creating robust importance sampling estimators via winsorization, with finite-sample optimality guarantees.

Feb 8, 2019 2:30 PM

Slides Code Preprint

Forecast Rodeo

Results and award-winning methods of a year-long data challenge to predict the weather 2-6 weeks in advance.

Oct 19, 2018 2:00 PM

Slides Code Preprint Poster Dataset News (USBR) News (Microsoft)

Scalable MCMC for Bayes Shrinkage Priors

How to scale one of the main Bayesian models for sparse high-dimensional regression to hundreds of thousands of predictors.

Jul 2, 2018 5:00 PM MCQMC 2018 Conference

Slides Preprint

Codes and Chains

Using Markov chains to decode ciphered messages written in graffiti across the walls of Rio de Janeiro.

Dec 3, 2012 10:00 AM VI Brazilian Mathematics Biennial

Slides Text News (Piauí)

Teaching

As an instructor at Stanford University:

STATS302: Qualifying Exams Workshop (Probability Theory). Summer 2017.

As a teaching assistant at Stanford University:

STATS315B: Modern Applied Statistics II (Graduate). Spring 2018.
STATS370: Bayesian Statistics (Graduate). Winter 2018.
STATS202: Data Mining and Analysis (Graduate). Fall 2015 and 2017.
MATH230B: Theory of Probability II (Graduate). Winter 2017.
MATH230A: Theory of Probability I (Graduate). Fall 2016.
STATS216: Statistical Learning (Graduate). Fall 2014, Summer 2015 and 2016.
STATS160: Statistical Methods (Undergraduate). Spring 2016.
STATS200: Statistical Inference (Graduate). Winter 2016.

As a teaching assistant at PUC-Rio:

MAT2621: Measure Theory and Integration (Graduate). Spring 2012.
ECO1215: Game Theory (Graduate). Fall 2011.
ECO 1113: Microeconomic Theory I (Undergraduate). Winter 2010.
ECO1721: Statistics and Econometrics (Undergraduate). Fall 2010.