Online Prediction in Sub-linear Space

Explore a groundbreaking Google TechTalk on online learning algorithms presented by Fred Zhang. Delve into the first sub-linear memory algorithm for online learning with expert advice, a fundamental concept in sequential decision making. Discover how this innovative approach overcomes the linear space complexity limitation of the traditional multiplicative weights update method while maintaining optimal regret. Learn about the main techniques, recent developments, and open research directions in this field. Gain insights from Fred Zhang, a fifth-year PhD student in Berkeley's theory group, as he discusses his joint work with Binghui Peng on this cutting-edge topic in algorithmic learning and statistics.