Speaker
-
Bernardino RomeraResearch Scientist en Google DeepMind
Alphatensor: descubriendo algoritmos de multiplicación de matrices usando aprendizaje por refuerzo
Bernardino Romera, Research Scientist at Google DeepMind, will be on December 15 in the ETSINF Assembly Hall, talking about #AlphaTensor, a reinforcement learning agent based on AlphaZero to discover efficient and demonstrably correct algorithms for matrix multiplication arbitrary.
Bernardino showed how they have turned the task of discovering such algorithms into a solitaire game that AlphaTensor can learn to play, beating new levels. He will also describe the key ingredients that make it possible to tackle these kinds of math problems using reinforcement learning.
