时 间： 2019年12月26日 13：00-14：00
地 点：中北校区理科大楼A1716报告厅
题 目：What actually matters in RL and possible (useful) research directions
主讲人：Wang Ran
主持人：吴贤毅教授
摘要：1. Recap on AlphaZero and AlphaStar.
2. Design of imitation learning networks. We will talk about VAE, Compositional Attention Network, Pointer Network, AutoML.
3. Asynchronous methods in RL (Why Q-learning sucks); some pitfalls.
4. Monte Carlo and Epistemic Logic.
报告人简介：Wang Ran graduated from Peking University with a bachelor's degree and studied at the University of Amsterdam with a master's degree in BI, mathematics, and econometrics. Among them, he obtained the highest GPA in mathematics and econometrics in the Netherlands.
Ran Worked at Percentage Technology Co., Ltd. from 2017 to 2019. From 2017 to 2018, he served as the project director of China Construction Bank's Data Analysis Center and participated in a large number of key projects across the bank, including small and micro fast loan line calculations, risk control cloud construction, intelligent dialogue robot optimization, and Dragon Fortune Portrait. The total number of projects undertaken during the period exceeded the sum of other vendors (IBM and Teradata).
From 2018 to 2019, he was in charge of the AI laboratory. He led the development of intelligent dialogue robots (similar to Baidu UNIT) and text proofreading products. Among them, the accuracy of text proofreading products is significantly higher than that of similar products in the industry. During the period, he also participated in numerous data mining competitions and achieved excellent results.
Wang Ran has always maintained a strong interest in reinforcement learning, as well as theoretical mathematics. On this matter, he has collaborated with numerous gaming companies, as well as MSRA and Prof. Zhihua Zhang.