Explore the world of online personalized recommendations through a deep dive into Deep Bayesian Bandits in this 29-minute video lecture. Delve into Contextual Bandit Theory, examining Upper Confidence Bound (UCB) and Thompson Sampling techniques. Gain insights into various approaches including Epsilon Greedy, bootstrapping, dropout method, and hybrid methods. Analyze experimental setups and results, considering factors such as time-to-click delay and fake negatives. Enhance your understanding of advanced recommendation algorithms and their practical applications in online personalization.
Deep Bayesian Bandits - Exploring in Online Personalized Recommendations
Overview
Syllabus
Introduction
Context
Bandit Theory
Upper Confidence Bound Ucb and Thompson Sampling
References
Approach
Experiments
Epsilon Greedy
Thompson Sampling
Upper Confidence Bound
bootstrapping
dropout method
hybrid method
upper confidence bounds
time to click delay
fake negatives
experiment setup
experiment results
Taught by
Launchpad
