Thursday, May 11 at 4:00pm to 5:00pm
925 E 13th Avenue, Eugene, OR 97403
Title: Neural networks and boolean functions.
Abstract: Deep neural networks can drive cars, produce images from text prompts, and write titles and abstracts for math lectures (J/K...or am I?). A major source of their power is their ability to approximate any function on a compact set. I'll tell you just enough about neural networks to convince you that this is possible, by focusing on a toy result: any boolean function can be modeled by a neural network of sufficiently high complexity.