Dylan Foster
I am a senior researcher at Microsoft Research, New England, where I am am member of the Reinforcement Learning Group. Previously, I was a postdoctoral fellow at the MIT Institute for Foundations of Data Science, and prior to this I received my PhD from the Department of Computer Science at Cornell University (2019), advised by Karthik Sridharan. I received my BS and MS in electrical engineering from the University of Southern California in 2014. I work on theory for machine learning.
Research
My research lies at the intersection of machine learning and decision making, including data-driven reinforcement learning and control, contextual bandits, and statistical learning in causal/counterfactual settings. I am interested in uncovering new algorithmic principles and fundamental limits for data-driven decision making, with themes including:
- Sample efficiency. What features of a class of models characterize the number of samples required to learn when we move from statistical learning to more demanding interactive tasks?
- Algorithm design. Can we reduce challenging data-driven decision making problems to classical machine learning primitives such as supervised classification and regression?
- Adaptivity. Can we develop data-driven decision making procedures that go beyond the worst case and adapt to favorable properties of real-world instances to get improved performance?
I am also excited about developing new models that better capture challenges and constraints faced when deploying data-driven decision making systems in practice.
More broadly, I am interested in all modern aspects of statistical learning, generalization, and algorithm design, especially in the context of deep learning and non-convex optimization.
Selected Awards
Recent News
| Dec 2020 | I gave a talk about the Instance-Dependent Complexity of Contextual Bandits and RL at the RL Theory Virtual Seminar. You can watch the recording here. |
| Oct 2020 | We are organizing a Spring 2022 program on Learning and Games at the Simons Institute. Apply for fellowships here. |
| Oct 2020 | Two new preprints I'm excited about: 1) Instance-Dependent Complexity of Contextual Bandits and RL, 2) Learning the LQR from Nonlinear Observations. |
| Aug 2020 | I am currently a (virtual) visitor at the Simons Institute for the Fall 2020 program on Foundations of Reinforcement Learning. |
| Mar 2020 | My talk about SquareCB at the 14th Annual New York Academy of Science ML Symposium received an award for best spotlight talk. |
| Nov 2019 | I am on the senior program committee for COLT 2020. |
| July 2019 | Received the Best Paper Award and Best Student Paper Award at COLT 2019. |
| Apr 2019 | In summer 2019 I will be a long-term visitor at the Simons Institute for the Foundations of Deep Learning program. |
| Jan 2019 | I have officially completed my PhD and started my postdoc at MIT! |
| Sep 2018 | I was featured on the Facebook research blog. |
| Aug 2018 | I am on the PC for ALT 2019. Submit your best work! |
| Jun 2018 | I received the Best Student Paper Award at COLT 2018 for Logistic Regression: The Importance of Being Improper. |
| Feb 2018 | In summer 2018 I will be an intern at MSR New England working with Vasilis Syrgkanis |
| Jan 2018 | I received the Facebook PhD Fellowship. |
| Sep 2017 | Two new papers at NIPS 2017 with spotlight presentations! |
| Apr 2017 | New paper ZigZag on online learning, decoupling, and Burkholder functions. |
| Feb 2017 | Spending summer 2017 as an intern with Rob Schapire and Miroslav Dudik at MSR NYC. |
| Jan 2017 | In Berkeley until May 2017 at the Simons Institute for the Foundations of Machine Learning program. |
Representative Papers (Show all):
Learning and Decision Making
Efficient First-Order Contextual Bandits:
Prediction, Allocation, and Triangular Discrimination
Dylan J. Foster and Akshay Krishnamurthy.
NeurIPS 2021. Oral presentation.Dylan J. Foster, Alexander Rakhlin, David Simchi-Levi, and Yunzong Xu
COLT 2021.Learning the Linear Quadratic Regulator from Nonlinear Observations
Zakaria Mhammedi, Dylan J. Foster, Max Simchowitz, Dipendra Misra,
Wen Sun, Akshay Krishnamurthy, Alexander Rakhlin, and John Langford.*
NeurIPS 2020.Beyond UCB: Optimal and Efficient Contextual Bandits
with Regression Oracles
Dylan J. Foster and Alexander Rakhlin.
ICML 2020.
Now in Vowpal Wabbit! Use the --squarecb option or see here for more information.Model Selection for Contextual Bandits
Dylan J. Foster, Akshay Krishnamurthy, and Haipeng Luo.
NeurIPS 2019. Spotlight presentation.
Practical Contextual Bandits with Regression Oracles
Dylan J. Foster, Alekh Agarwal, Miroslav Dudík, Haipeng Luo, and Robert E. Schapire.*
ICML 2018. Long talk.
Now in Vowpal Wabbit (thanks to Alberto Bietti!). Try it with the --regcb or --regcbopt option.Logistic Regression: The Importance of Being Improper
Dylan J. Foster, Satyen Kale, Haipeng Luo, Mehryar Mohri, and Karthik Sridharan.
COLT 2018. Best Student Paper Award.
Online Learning: Sufficient Statistics and the Burkholder Method
Dylan J. Foster, Alexander Rakhlin, and Karthik Sridharan.
COLT 2018.
Dylan J. Foster, Alexander Rakhlin, and Karthik Sridharan.
NeurIPS 2015. Spotlight presentation.
Modern Challenges in Statistical Learning
Statistical Learning with a Nuisance Component
Dylan J. Foster and Vasilis Syrgkanis.
COLT 2019. Best Paper Award.
Extended abstract for Orthogonal Statistical Learning (reviewed as full paper).
Lower Bounds for Non-Convex Stochastic Optimization
Yossi Arjevani, Yair Carmon, John C. Duchi, Dylan J. Foster,
Nathan Srebro, and Blake Woodworth.
Mathematical Programming, Series A, 2022.The Complexity of Making the Gradient Small in
Stochastic Convex Optimization
Dylan J. Foster, Ayush Sekhari, Ohad Shamir, Nathan Srebro, Karthik Sridharan,
and Blake Woodworth.
COLT 2019. Best Student Paper Award.
Spectrally-Normalized Margin Bounds for Neural Networks
Peter Bartlett, Dylan J. Foster, and Matus Telgarsky.
NeurIPS 2017. Spotlight presentation.
Thesis
Adaptive Learning: Algorithms and Complexity
Dylan J. Foster
Ph.D. Thesis. Department of Computer Science, Cornell University, 2019.
Cornell CS Doctoral Dissertation Award.
Service
Program Committee/Area Chair: COLT (Senior PC): 2020, 2021, NeurIPS (Area Chair): 2020, ALT: 2019, 2020, 2021, Learning for Dynamics and Control (L4DC): 2020, 2021.
Conference Reviewing: COLT, NeurIPS, ICML, STOC, FOCS, SODA, ALT, AISTATS, AAAI.
Journal Reviewing: JMLR, Journal of the ACM, Annals of Statistics, Mathematics of Operations Research, Operations Research, Biometrika.
Teaching
Machine Learning Theory
Cornell University, Spring 2018.
Professor Karthik Sridharan.
Introduction to Analysis of Algorithms
Cornell University, Spring 2015.
Professors Éva Tardos and David Steurer.
Received outstanding teaching award.Foundations of Artificial Intelligence
Cornell University, Fall 2014.
Professor Bart Selman.
Contact
I can be reached at dylanfoster at microsoft dot com.
© Dylan Foster 2015.