Y LeCun
Predictive Learning
Yann Le Cun
Facebook AI Research
Center for Data Science, NYU
Courant Institute of Mathematical Sciences, NYU
http://yann.lecun.com
Y LeCun
1st Connectionist Summer School, CMU July 1986
Y LeCun
1st Connectionist Summer School, CMU July 1986
Jay McClelland Terry Sejnowski
Geoff Hinton
Mike Jordan
Jordan Pollack
Jim Hendler
Dave Touretzky
Yann LeCun
Richard Durbin
Bartlett Mel
Stan Dehaene
Andy Barto
Mike Mozer
Dana Ballard
Gerry Tesauro
David Willshaw
Bart Selman
Y LeCun
Supervised Learning
We can train a machine on lots of examples of tables, chairs,
dog, cars, and people
But will it recognize table, chairs, dogs, cars, and people it has
never seen before?
CAR
PLANE
CAR
Y LeCun
Deep Learning = The Entire Machine is Trainable
Traditional Pattern Recognition: Fixed/Handcrafted Feature Extractor
Trainable
Classifier
Feature
Extractor
Mainstream Modern Pattern Recognition: Unsupervised mid-level features
Trainabl
Feature/Extractor/ble/Traina/Learning/Pattern/chairs/LeCun/cars/people/
Feature/Extractor/ble/Traina/Learning/Pattern/chairs/LeCun/cars/people/
-->