Explore a 44-minute lecture by Surya Ganguli from Stanford University, presented at the Simons Institute, delving into the exact solutions of geometric dynamics in signal propagation through transformers. Gain insights into how these solutions predict transformer behavior and contribute to understanding transformers as a computational model. Examine the intricate mathematical foundations underlying transformer architectures and their implications for signal processing and information flow in deep learning systems.
Exact Solutions to the Geometric Dynamics of Signal Propagation Through Transformers
Overview
Syllabus
Exact solutions to the geometric dynamics of signal propagation through transformers predict...
Taught by
Simons Institute
Related Courses
-
How Do Transformers Work? - A Deep Dive into Neural Network Architecture
-
Introduction to Transformers
-
The Emergence of Clusters in Self-Attention Dynamics
-
Large Language Models and Transformers: Perspectives from Physics, Neuroscience, and Theory
-
Transformers as a Computational Model
-
Toward Length Extrapolatable Transformers
