François Charton | Transformers for maths, and maths for transformers
HTML-код
- Опубликовано: 22 авг 2024
- New Technologies in Mathematics Seminar 9/27/2023
Speaker: François Charton, Meta AI
Title: Transformers for maths, and maths for transformers
Abstract: Transformers can be trained to solve problems of mathematics. I present two recent applications, in mathematics and physics: predicting integer sequences, and discovering the properties of scattering amplitudes in a close relative of Quantum ChromoDynamics.
Problems of mathematics can also help understand transformers. Using two examples from linear algebra and integer arithmetic, I show that model predictions can be explained, that trained models do not confabulate, and that carefully choosing the training distributions can help achieve better, and more robust, performance.
Curious if anyone on this channel is familiar with my great-grandfather, Professor Bartlett Jere Whiting. I’ve read his work except for “When Evensong and Morrowsong Accord: Three Essays on the Proverb”. I’d like to know if there is anything written that speaks to his personality or perspectives.