Distilled Thompson Sampling: Practical and Efficient Thompson Sampling via Imitation Learning
Hongseok Namkoong, Sam Daulton, Eytan Bakshy
I am a Research Scientist in the Core Data Science team’s Adaptive Experimentation group. My research focuses on developing methods for contextual bandit optimization, reinforcement learning, and Bayesian optimization. Prior to joining Meta, I was at Harvard University, where my work focused on developing robust and efficient transfer learning methods to advance towards disseminating reinforcement learning to high-stakes human applications.
Interests
Bayesian optimization, contextual bandits, reinforcement learning, transfer learning
Latest Publications
Offline Reinforcement Learning Workshop at NeurIPS - 2020
Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective Bayesian Optimization
Sam Daulton, Max Balandat, Eytan Bakshy
Parallel Bayesian Optimization of Multiple Noisy Objectives with Expected Hypervolume Improvement
Sam Daulton, Max Balandat, Eytan Bakshy
Optimizing Coverage and Capacity in Cellular Networks using Machine Learning
Ryan M. Dreifuerst, Sam Daulton, Yuchen Qian, Paul Parayil Varkey, Max Balandat, Sanjay Kasturia, Anoop Tomar, Ali Yazdan Panah, Vish Ponnampalam, Robert W. Heath Jr
Latency-Aware Neural Architecture Search with Multi-Objective Bayesian Optimization
David Eriksson, Pierce I-Jen Chuang, Sam Daulton, Peng Xia, Akshat Shrivastava, Arun Babu, Shicong Zhao, Ahmed Aly, Ganesh Venkatesh, Max Balandat
Thompson Sampling for Contextual Bandit Problems with Auxiliary Safety Constraints
Sam Daulton, Shaun Singh, Vashist Avadhanula, Drew Dimmery, Eytan Bakshy
