I'm a principal scientist and director on the Core Data Science Team at Meta, where I lead the Adaptive Experimentation team. I am particularly interested developing practical and robust methods for sequential experimentation and reinforcement learning for real-world applications. Much of my work has been motivated by the use of randomized experiments for understanding social behavior, including influence and information diffusion in networks.
Here are a few of my recent publications. For a more complete list of publications and activities, see my Google Scholar page and CV.
Bayesian optimization and reinforcement learning
- Robust Multi-Objective Bayesian Optimization Under Input Noise.
Samuel Daulton, Sait Cakmak, Maximilian Balandat, Michael A Osborne, Enlu Zhou, Eytan Bakshy. Accepted at ICML 2022.
- Multi-Objective Bayesian Optimization over High-Dimensional Search Spaces.
Samuel Daulton, David Eriksson, Maximilian Balandat, Eytan Bakshy. Accepted at UAI 2022.
- Preference Exploration for Efficient Bayesian Optimization with Multiple Outcomes.
Jerry Lin, Raul Astudillo, Peter Frazier, Eytan Bakshy. AISTATS 2022.
- Look-Ahead Acquisition Functions for Bernoulli Level Set Estimation.
Benjamin Letham, Eytan Bakshy, Michael Shvartsman. AISTATS 2022.
- Bayesian Optimization with High-Dimensional Outputs.
Wesley Maddox, Max Balandat, Andrew Gordon Wilson, Eytan Bakshy. NeurIPS 2021.
Multi-Step Budgeted Bayesian Optimization with Unknown Evaluation Costs.
Raul Astudillo, Daniel Jiang, Maximilian Balandat, Eytan Bakshy, Peter Frazier. NeurIPS 2021.
- Parallel Bayesian Optimization of Multiple Noisy Objectives with Expected Hypervolume Improvement.
Samuel Daulton, Max Balandat, Eytan Bakshy. NeurIPS 2021.
- Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective Bayesian Optimization.
Samuel Daulton, Max Balandat, Eytan Bakshy. NeurIPS 2020.
- Re-Examining Linear Embeddings for High-Dimensional Bayesian Optimization.
Benjamin Letham, Roberto Calandra, Akshara Rai, Eytan Bakshy. NeurIPS 2020.
- BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization.
Maximilian Balandat, Brian Karrer, Daniel R. Jiang, Samuel Daulton, Benjamin Letham, Andrew Gordon Wilson, Eytan Bakshy. NeurIPS 2020.
- High-Dimensional Contextual Policy Search with Unknown Context Rewards using Bayesian Optimization.
Qing Feng, Benjamin Letham, Hongzi Mao, Eytan Bakshy. NeurIPS 2020.
- Distilled Thompson Sampling: Practical and Efficient Thompson Sampling via Imitation Learning.
Sam Daulton, Hongseok Namkoong, Eytan Bakshy. NeurIPS 2020 Offline Reinforcement Learning Workshop.
- Preference Learning for Real-World Multi-Objective Decision Making.
Jerry Lin, Eytan Bakshy. ICML 2020 Workshop on Real World Experiment Design and Active Learning.
- Bayesian Optimization for Policy Search via Online-Offline Experimentation.
Ben Letham & Eytan Bakshy. JMLR 2019.
- Constrained Bayesian Optimization with Noisy Experiments.
Benjamin Letham, Brian Karrer, Guilherme Ottoni, Eytan Bakshy. Bayesian Analysis 2019.
- Thompson Sampling for Contextual Bandit Problems with Auxiliary Safety Constraints.
Samuel Daulton, Shaun Singh, Vashist Avadhanula, Drew Dimmery, Eytan Bakshy. NeurIPS 2019 Workshop of Safe Reinforcement Learning.
- Real-world Video Adaptation with Reinforcement Learning.
Hongzi Mao, Shannon Chen, Drew Dimmery, Shaun Singh, Drew Blaisdell, Yuandong Tian, Mohammad Alizadeh, Eytan Bakshy.
ICML 2019 Workshop on Reinforcement Learning for Real Life.
- AE: A domain-agnostic platform for adaptive experimentation.
Eytan Bakshy, Lili Dworkin, Brian Karrer, Konstantin Kashin, Ben Letham, Ashwin Murthy, Shaun Singh. NeurIPS 2018 Systems for ML Workshop.
- Scalable Meta-Learning for Bayesian Optimization.
Matthias Feurer, Benjamin Letham, Eytan Bakshy. ICML 2018 AutoML Workshop.
Causal Inference and Experimentation
- PlanAlyzer: assessing threats to the validity of online experiments.
Emma Tosch, Eytan Bakshy, Emery D Berger, David D Jensen, J Eliot B Moss. Communications of the ACM, 2021.
- Bias and High-Dimensional Adjustment in Observational Studies of Peer Effects.
Dean Eckles & Eytan Bakshy. Journal of the American Statistical Association, 2020.
- Shrinkage Estimators in Online Experiments.
Drew Dimmery, Eytan Bakshy, Jasjeet Sekhon. KDD 2019.
- Design and Analysis of Benchmarking Experiments for Distributed Internet Services.
Eytan Bakshy & Eitan Frachtenberg. WWW 2015.
- Designing and Deploying Online Field Experiments.
Eytan Bakshy, Dean Eckles, Michael Bernstein. WWW 2014.
- Uncertainty in Online Experiments with Dependent Data: An Evaluation of Bootstrap Methods.
Eytan Bakshy & Dean Eckles. KDD 2013.
- Social Influence and Political Mobilization: Further Evidence from a Randomized Experiment in the 2012 U.S. Presidential Election.
Jason Jones, Robert Bond, Eytan Bakshy, Dean Eckles, James Fowler. PLoS ONE, 2017.
- Estimating Peer Effects with Peer Encouragement Designs.
Dean Eckles, Rene Kizilcec, Eytan Bakshy. PNAS 2016.
- Selection Effects in Online Sharing: Consequences for Peer Adoption.
Sean J. Taylor, Eytan Bakshy, Sinan Aral. EC 2013.
- Social Influence in Social Advertising: Evidence from Field Experiments.
Eytan Bakshy, Dean Eckles, Rong Yan, Itamar Rosenn. EC 2012.
- The Role of Social Networks in Information Diffusion.
Eytan Bakshy, Itamar Rosenn, Cameron A. Marlow, Lada A. Adamic. WWW 2012.
Exposure to Ideologically Diverse News and Opinion on Facebook.
Eytan Bakshy, Solomon Messing, Lada Adamic. Science, 2015.
Additional materials: Technical notes on paper and future research, replication materials and additional data.
- Quantifying the Invisible Audience in Social Networks.
Michael Bernstein, Eytan Bakshy, Moira Burke, Brian Karrer. CHI 2013.
- Center of Attention: How Facebook Users Allocate Attention across Friends.
Lars Backstrom, Eytan Bakshy, Jon Kleinberg, Thomas M. Lento, Itamar Rosenn. ICWSM 2011.
- Everyone's an Influencer: Quantifying Influence on Twitter.
Eytan Bakshy, Jake M. Hofman, Winter A. Mason, Duncan J. Watts. WSDM 2011.
- Social Influence and the Diffusion of User-Contributed Content
Eytan Bakshy, Brian Karrer, Lada A. Adamic. EC 2009.
The Adaptive Experimentation team at Meta develops two open-source software packages:
- Ax: A domain-agnostic platform for adaptive experimentation. Ax powers AutoML at Meta, A/B-test-based parameter tuning experiments, infrastructure optimization, hardware design, perception research, and robotics research.
- BoTorch: A library for Bayesian optimization research. It uses an extremely modular design and closely integrates with (G)PyTorch to enable state-of-the-art rersearch that combines deep Bayesian models and Bayesian optimization. It includes extensive support for Monte Carlo acquisition functions, automatic gradients, and multi-objective optimization.
I am also the creator of PlanOut. It's an open-source framework for designing and implementing complex behavioral science experiments.
I find PlanOut to be helpful for thinking about and planning experiments, and hope that people can use it to run interesting experiments.
Visit the PlanOut homepage and give it a try.
For more on designing and deploying online field experiments with PlanOut, check out our paper on the subject.
Coverage on papers and Data Science Team posts:
Exposure to Ideologically Diverse Information on Facebook
Coverage on original paper @ New York Times, Science Magazine News, Christian Science Monitor
- Big experiments: Big data's friend for making decisions.
Coverage @ Junk Charts / Numbersense
- Showing Support for Marriage Equality on Facebook.
Coverage @ AllThingsD, The Atlantic, Gizmodo, Wall St. Journal, Forbes, CNN, BBC News
- Estimating Audience Size on Facebook
- The 2012 Election through the Facebook Lense
Coverage @ MIT Technology Review, Talking Points Memo
- Rethinking Information Diversity in Networks
Coverage @ Slate, New Scientist