Please use this zoom link for the event.
Please join us for a UW Data Science Seminar event on Thursday, January 27th from 4:30 to 5:30 p.m. The seminar will feature Sareh Nabi, a postdoctoral scholar in the Amazon Advertising group.
“Bayesian Meta-Prior Learning Using Empirical Bayes”
Abstract: Adding domain knowledge to a learning system is known to improve results. In multiparameter Bayesian frameworks, such knowledge is incorporated as a prior. On the other hand, the various model parameters can have different learning rates, especially with skewed data. Two often-faced challenges in real-world applications are the absence of informative priors and the inability to control parameter learning rates. In collaboration with my co-authors, we proposed a hierarchical empirical Bayes approach that addresses both challenges and that can generalize to any Bayesian framework (link here). Our method learns empirical meta-priors from initial stream of data and uses them to decouple the learning rates of first-order and second-order features (or any other given feature grouping) in a generalized linear model. In this talk, I will review our approach and theoretical results for the unbiasedness, strong consistency, and optimal frequentist cumulative regret properties of our meta-prior variance estimator. In the last part of my talk, I’ll also present both our simulations and live experiments in Amazon production system that demonstrated marked improvements, especially in cases of small traffic that are often a challenge.
Biography: Sareh Nabi is a postdoctoral scholar in the sponsored ads team at Amazon, working under Dr. Lihong Li’s supervision and in close collaboration with Prof. Marciano Siniscalchi and Dr. Ming Chen. Sareh joined the Amazon Advertising group as part of the early career scientist program launched in the spring of 2021. Her current research at Amazon focuses on multi-agent Reinforcement Learning (RL) applications in advertising. Before Amazon, Sareh worked at Microsoft for a few years, focused on integrating machine learning solutions with MS Dynamics business products. She received her Ph.D. in Information System and Operations Management from the University of Washington in 2018 under Prof. Hamed Mamani’s supervision. Her Ph.D. research focused on the applications of contextual bandits in dynamic pricing and online advertising. Her previous educational background is an undergrad in Math from the Sharif University of Technology and dual masters in Math and Economics from Simon Fraser University.
The UW Data Science Seminar is an annual lecture series at the University of Washington that hosts scholars working across applied areas of data science, such as the sciences, engineering, humanities and arts along with methodological areas in data science, such as computer science, applied math and statistics. Our presenters come from all domain fields and include occasional external speakers from regional partners, governmental agencies and industry.
The 2021-2022 seminars will be virtual events, and are free and open to the public.