Dear colleagues,
we would like to invite you to the seminar of Division of Elementary Particle Physics of Institute of Physics, presented by .
For more info, please see invitation.
Abstract:
Deep learning architectures designed for natural language, such as the transformer, can be used to solve problems of mathematics and physics, by considering the problems and solutions as sequences of words into some formal language, and training the model to translate, from examples only, the problem into the solution. Conversely, problems of mathematics often prove to be good benchmarks for understanding. I will present applications of transformers to advanced mathematical problems, and results on mathematical benchmarks that help understand how language models learn.
