BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Cambridge ELLIS Seminar Series- Dr Roger Grosse- Studying Neural N
 et Generalization through Influence Functions - Speaker to be confirmed
DTSTART:20230306T161500Z
DTEND:20230306T171500Z
UID:TALK197914@talks.cam.ac.uk
CONTACT:Kimberly Cole
DESCRIPTION:The Cambridge ELLIS Unit Seminar Series holds talks by leading
  researchers in the area of machine learning and AI. Next first  speaker o
 f 2023 will be Dr. Roger Grosse. Details of his talk can be found below. 
 \n\nTitle: “Studying Neural Net Generalization through Influence Functio
 ns"\n\nAbstract: How can we trace surprising behaviors of machine learning
  models back to their training data?  Influence functions aim to predict 
 how the trained model would change if a specific training example were add
 ed to the training set. I’ll address two issues that have blocked their 
 applicability to large-scale neural nets: apparent inaccuracy of the resul
 ts\, and the difficulty of computing inverse-Hessian-vector products. Towa
 rds the former issue\, I’ll reformulate the goals of influence estimatio
 n in a way that applies to overparameterized\, incompletely trained models
 \, and argue that the apparent inaccuracy was largely illusory. I’ll the
 n discuss an approach to scaling influence estimation to large language mo
 dels and show some resulting insights into their patterns of generalizatio
 n..\n\nhttps://eng-cam.zoom.us/j/82557742057?pwd=Sk1sRWRxL09pV2FPYzNpK0Fr
 U096Zz09\n
LOCATION:https://eng-cam.zoom.us/j/82557742057?pwd=Sk1sRWRxL09pV2FPYzNpK0F
 rU096Zz09
END:VEVENT
END:VCALENDAR