[talks] J Deng preFPO

[Melissa M. Lawson]

Jia Deng will present his preFPO on Monday January 9 at 1:30 PM in Room 402.
The members of his committee are:  Fei-Fei Li, advisor; Kai Li and Szymon 
Rusinkiewicz, readers; David Blei and Adam Finkelstein, nonreaders.  Everyone 
is invited to attend his talk.  His abstract follows below.
-----------------

Title: Building the forest to see trees: toward large-scale visual recognition

Abstract:
Visual recognition is inherently a large-scale problem in that humans
can recognize tens of thousands of categories. One of the goals of
computer vision is to endow machines with this ability. In this talk,
I will discuss recent advances toward recognition at this human scale.
I will first talk about the acquisition and analysis of large-scale data,
in particular, the construction of a large-scale visual ontology and
an evaluation of state of the art recognition algorithms on more than
10,000 categories. Then I will present two recent works on large-scale
machine learning. The first is on speeding-up training and testing to
achieve sub-linear costs with respect to the number of categories. The
second is about optimizing accuracy-specificity trade-offs to achieve
reliable multi-level categorization. I will demonstrate that, by
exploiting the structures among categories, our methods significantly
improve the computational efficiency and reliability of recognition
over previous work.