The AI Innovation Institute (AI3) is set to host a seminar featuring Professor Meir Feder from Tel-Aviv University on Monday, November 10. The event, titled “Information-Theoretic Framework for Understanding Modern Machine-Learning,” will take place at 12:30 pm in the New Computer Science Building, Room 120.
The seminar will focus on how Information Theory interprets learning as universal prediction under log loss and examines this process through regret bounds. The approach discussed by Professor Feder differs from traditional methods that dealt with smaller model classes and offered uniform regret. Instead, his framework provides non-uniform, model-dependent bounds based on architecture-driven measures of model complexity.
Professor Feder holds the Jokel Chair in Information Theory at Tel-Aviv University’s School of Electrical and Computer Engineering. He has been recognized with several honors, including the IEEE Information Theory Society best paper award and the Padovani lectureship. In addition to his academic achievements, he has founded five companies such as Peach Networks and Amimon.
With a renewed interest in machine learning and artificial intelligence, Professor Feder co-founded Run:ai—a company focused on virtualization and orchestration platforms for AI infrastructure—which was later acquired by Nvidia to enhance its GPU cloud operations.
His work at Amimon led to significant recognition; he received the 2020 Scientific and Engineering Award from the Academy of Motion Picture Arts and Sciences. He was also named principal inventor of technology that won the 73rd Engineering Emmy Award from the Television Academy.
"Information Theory views learning as universal prediction under log loss, characterized through regret bounds. Unlike the classical results that considered 'small' model classes and provided uniform regret, the proposed framework provides non-uniform, model dependent bounds utilizing an effective notion of architecture-based model complexity," said organizers about Professor Feder's research focus.