[Colloquium] 3/28 Talks at TTIC: Zaid Harchaoui, NYU

[Mary Marre]

When:     Monday, March 28th at 11:00 am

Where:    TTIC, 6045 S Kenwood Avenue, 5th Floor, Room 526

Who:       Zaid Harchaoui, NYU

Title:       Towards Deep Convolutional Methods without Supervision

Abstract:
We introduce a new kernel-based approach for the design of multi-layer
convolutional feature representations for signals and images. The majority
of deep convolutional neural networks (ConvNets) require supervision to
learn state-of-the-art feature representations, which can be problematic
for real-world applications where supervision can be expensive or
unconceivable. Unsupervised learning of feature representations is
currently a major open problem.

The proposed approach allow to define a feature representation that is
based on a kernel (feature) map. The exact version of this feature
representation is data-independent. An explicit kernel (feature) map can be
computed to approximate it for computational efficiency, without the need
for supervision. We illustrate the potential of the approach on several
applications: i) patch matching; ii) image retrieval; iii) image
classification. We settle new state-of-the-art, or compete on par with
state-of-the-art methods, on public benchmarks for all three computer
vision tasks.


Bio:
Zaid Harchaoui is currently Visiting Assistant Professor at the Courant
Institute for Mathematical Sciences of NYU. Zaid was a permanent researcher
at Inria from 2010 to 2015. He received his PhD from ParisTech (Paris,
France). He received the Inria award for scientific excellence and the NIPS
reviewer award. He gave a tutorial on "Frank-Wolfe, greedy algorithms, and
friends" at ICML'14, on "Large-scale visual recognition" at CVPR'13, and on
"Machine Learning for Computer Vision" at MLSS Kyoto 2015. He recently
co-organized the workshop on “Optimization for Machine Learning” at
NIPS'14, and the "Optimization and Statistical Learning" workshop in 2015
and 2013 in Ecole de Physique des Houches (France). He is associate editor
of IEEE Signal Processing Letters. He is Area Chair for ICML 2015, ICML
2016, and NIPS 2016.



Host: Nathan Srebro, nati at ttic.edu





Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 504*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20160323/3f6962c4/attachment.htm