Research and Publications

My research is broadly in the areas of optimization and machine learning. A lot of my work has been dedicated to developing new methodologies and foundational theory of operations-driven machine learning, which is concerned with developing better decision-focused machine learning models for predicting unknown quantities associated with an optimization model used for decision-making. I also work a lot on developing and analyzing large-scale first-order optimization methods, with particular emphasis on structure-enhancing algorithms, connections to popular algorithms in statistics, and applications to data-driven decision-making and machine learning problems. My research has been generously funded by the National Science Foundation (NSF) and UC Berkeley.

Videos of some recent talks:

Offline and Online Learning for Contextual Stochastic Optimization at IPAM UCLA, where I gave an overview of some of my work on Smart Predict-then-Optimize (SPO) and Integrated Conditional Estimation-Optimization (ICEO).
Learning, Optimization, and Generalization in the Predict-then-Optimize Setting at CPAIOR 2022, where I gave an overview of some of my work on Smart Predict-then-Optimize (SPO) and some extensions to the online decision-making setting.

Working Papers

Binary Classification with Instance and Label Dependent Label Noise, with Hyungki Im.
Active Learning in the Predict-then-Optimize Framework: A Margin-Based Approach, with Mo Liu, Heyuan Liu, and Zuo-Jun Max Shen.
Online Contextual Decision-Making with a Smart Predict-then-Optimize Method, with Heyuan Liu.
New Methods for Parametric Optimization via Differential Equations, with Heyuan Liu.
New Penalized Stochastic Gradient Methods for Linearly Constrained Strongly Convex Optimization, with Meng Li and Alper Atamturk.
Integrated Conditional Estimation-Optimization, with Meng Qi and Zuo-Jun Max Shen.
Optimal Bidding, Allocation and Budget Spending for a Demand Side Platform With Generic Auctions, with Alfonso Lobos, Zheng Wen, and Kuang-chih Lee.
- Preliminary extended abstract version was presented at the 2018 AdKDD & TargetAd Workshop at KDD, London, United Kingdom, 2018.
- Winner of the Best Student Paper Prize (Alfonso Lobos) at the 2018 AdKDD & TargetAd Workshop at KDD, London, United Kingdom, 2018.

Publications

Stochastic First-Order Algorithms for Constrained Distributionally Robust Optimization, with Hyungki Im, INFORMS Journal on Computing, forthcoming. [code]
On the Softplus Penalty for Large-Scale Convex Optimization, with Meng Li and Alper Atamturk, Operations Research Letters 51 (6), pp. 666-672, 2023.
Generalization Bounds in the Predict-then-Optimize Framework, with Othman El Balghiti, Adam N. Elmachtoub, and Ambuj Tewari, Mathematics of Operations Research 48 (4), pp. 2043-2065, 2023.
Ch3MS-RF: a random forest model for chemical characterization and improved quantification of unidentified atmospheric organics detected by chromatography–mass spectrometry techniques, with Emily B. Franklin, Lindsay D. Yee, Bernard Aumont, Robert J. Weber, and Allen H. Goldstein, Atmospheric Measurement Techniques 15 (12), pp. 3779-3803, 2022.
Smart “Predict, then Optimize,” with Adam N. Elmachtoub, Management Science 68 (1), pp. 9-26, 2022. [code]
- 1st place in the 2020 INFORMS Junior Faculty Interest Group (JFIG) Paper Competition.
- Featured article in Management Science Review, March 2022
- 20 minute video with Adam.
Risk Bounds and Calibration for a Smart Predict-then-Optimize Method, with Heyuan Liu, Advances in Neural Information Processing Systems (NeurIPS) 34, 2021.
Joint Online Learning and Decision-making via Dual Mirror Descent, with Alfonso Lobos and Zheng Wen, Proceedings of the 38th International Conference on Machine Learning (ICML) PMLR 139:7080-7089, 2021.
Generalization Bounds in the Predict-then-Optimize Framework, with Othman El Balghiti, Adam N. Elmachtoub, and Ambuj Tewari, Advances in Neural Information Processing Systems (NeurIPS) 32, pp. 14389-14398, 2019.
A New Perspective on Boosting in Linear Regression via Subgradient Optimization and Relatives, with Robert M. Freund and Rahul Mazumder, The Annals of Statistics 45 (6), pp. 2328-2364, 2017.
- Selected to be presented in the Annals of Statistics Special Invited Session at JSM 2017, which features the four best papers accepted to The Annals of Statistics in the previous two years.
- Winner of the 2015 INFORMS Optimization Society Student Paper Prize.
- Short article appeared in the 2016 INFORMS Optimization Society Newsletter.
An Extended Frank-Wolfe Method with “In-Face” Directions, and its Application to Low-Rank Matrix Completion, with Robert M. Freund and Rahul Mazumder, SIAM Journal on Optimization 27 (1), pp. 319-346, 2017. [code]
New Analysis and Results for the Frank-Wolfe Method, with Robert M. Freund, Mathematical Programming 155 (1), pp. 199-230, 2016.

Refereed Workshop Proceedings

Profit Maximization for Online Advertising Demand-Side Platforms, with Alfonso Lobos, Zheng Wen, and Kuang-chih Lee, Proc. AdKDD & TargetAd Workshop at KDD 2017, Halifax, Canada.
Incremental Forward Stagewise Regression: Computational Complexity and Connections to LASSO, with Robert M. Freund and Rahul Mazumder, Proc. International Workshop on Advances in Regularization, Optimization, Kernel Methods and Support Vector Machines (ROKS), Leuven, Belgium, 2013.

Technical Reports

Stochastic In-Face Frank-Wolfe Methods for Non-Convex Optimization and Sparse Neural Network Training, with Alfonso Lobos and Nathan Vermeersch, 2019.
Condition Number Analysis of Logistic Regression, and its Implications for Standard First-Order Solution Methods, with Robert M. Freund and Rahul Mazumder, 2018.
AdaBoost and Forward Stagewise Regression are First-Order Convex Optimization Methods, with Robert M. Freund and Rahul Mazumder, MIT Operations Research Center working paper OR 397-14, 2013.