I am member of Globo.com Big Data team responsible for promoting data-driven culture, as well as maintaining and coding new features for the A/B testing and MAB platform. As far as A/B test is concerned, we working with statistical hypothesis testing and statistic distributions. Concerning MAB, we have being applying reinforcement learning algorithm to optimize the best option choice (alternatives), using algorithms such as UCB1, UCBTuned, Thompson Sampling and Ranked MAB), applying Monte Carlo Simulation to understand their behaves. A/B and MAB Platform process more than 400 million records per day and supply a REST API to handle more than 40K RPS.
Evento | Trilha | Data | Palestra |
---|---|---|---|
TDC 2019 Recife | TRILHA INTELIGÊNCIA ARTIFICIAL E MACHINE LEARNING | 10/10/2019 | Reinforcement Learning na globo.com: Otimizando o melhor momento para carregar peças publicitárias |