EPFL AI Center x SDSC - Research Seminar Series - Alhussein Fawzi

The workshop is jointly organized by the EPFL AI Center and the Swiss Data Science Center (SDSC).

Title
Machine learning for discovering new algorithms and solving mathematical problems

Abstract 
In the modern era of computing, algorithms for solving fundamental problems such as computing the shortest path in graphs and solving linear equations, are used billions of times every day. However, paradoxically, such algorithms which are central to modern computing were often designed well before the advent of modern computation. In fact, discovering new efficient algorithms is notoriously difficult, and often involves solving prohibitively large combinatorial problems. In this talk, I will describe how to use machine learning techniques - in particular reinforcement learning and large language models - to discover new algorithms and solve long-standing mathematical problems. We will particularly focus on designing new algorithms for fundamental computational problems, such as matrix multiplication.

Bio
Alhussein Fawzi is a research scientist at Google DeepMind. He was awarded the MIT Innovators Under 35 award in 2023 for his work on discovering new algorithms with machine learning. He works on AI for Science, and is particularly interested in using Machine Learning to unlock new results in Computer Science, Algorithms, and Mathematics. He recently published two papers in Nature magazine, where he used Machine Learning and Large Language Models to discover new algorithms for fundamental computational tasks, and new mathematical results. Prior to working at Google DeepMind, he obtained his PhD from EPFL from the signal processing laboratory in 2016, working on the robustness and reliability of machine learning models.