Boosting and Low-rank

Monday, June 24, 2013 - 9:00am - 10:30am
Lind 305
Bin Yu (University of California, Berkeley)
Boosting is one of the two most successful machine learning methods
with SVM. It uses gradient descent to an empirical loss function.
When the step sizes are small, it is computationally efficient way
to approximate Lasso. When a nuclear norm penalization is applied to L2
we have the low-rank regularization arising from the Netflix competition.
A subset of the netflix data will be investigated.
MSC Code: