Speaker
Prof.
Marco Romito
(Università di Pisa)
Description
The first part of the talk will introduce a mathematical framework for transformers architectures, focusing in particular on how information travels through the layers of the network. In the second part we will briefly discuss infinite-width, infinite-depth and infinite-tokens limits. Finally, the third part will focus on synchronization and some noise-induced phenomena.
Co-authors
Prof.
Andrea Agazzi
(Università di Berna)
Eloy Mosig
(Università di Pisa)
Dr
Giuseppe Bruno
(Università di Berna)
Prof.
Marco Romito
(Università di Pisa)
Samuele Saviozzi
(Università di Pisa)