Deep learning methods in vision CVPR Tutorial Deepin Learning Methods for Vision

Deep learning Methods for Vision CVPR Tutorial 9:00am-5:30pm, Sunday June 17th, Ballroom D (full day) Rob Fergus (NYU), Honglak Lee (Michigan), Marc ' Aurelio Ranzato (Google) Ruslan Salakhutdinov (Toronto), Graham Taylor (Guelph), Kai Yu (Baidu)

Overview Hand-designed features such as SIFT and HOG underpin many successful object recognition approaches. However, these only capture low-level edge information and it have proven difficult to design features that effectively cap Ture mid-level cues (e.g. edge intersections) or high-level representation (e.g. object parts). However, recent developments in machine learning, known as "deep Learning", with shown how hierarchies of features can Learned in a unsupervised manner directly from data. This tutorial would describe these feature learning approaches, as applied to images and video. The tutorial would start by motivating the need to learn features, rather than hand-craft them. It would then introduce several basic architectures, explaining how they learn features, and showing how they can is "stack Ed "into hierarchies that can extract multiple layers of representation. Throughout, links would be drawn between these methods and existing approaches to recognition, particularly those involving Hierarchical representations. The final part of the lecture would examine the current performances obtained by feature learning approaches on a range of Standard vision benchmarks, highlighting their strengths and weaknesses.

Schedule

9:00am	Introduction Ppt	(Fergus)	[1h]
10:00am	Coffee break		[30m]
10:30am	Sparse Coding Ppt	(Yu)	[1h]
11:30am	Neural Networks PDF Code	(Ranzato)	[1h]
12:30pm	Lunch		[1h]
1:30pm	Restricted Boltzmann Machines Pdf	(Lee)	[1h]
2:30pm	Deep Boltzmann Machines Pdf	(Salakhutdinov)	[30m]
3:00pm	Coffee break		[30m]
3:30pm	Transfer Learning Pdf	(Salakhutdinov)	[30m]
4:00pm	Motion and Video Pdf	(Taylor)	[1h]
5:00pm	Summary/q & A		[30m]

Speaker Biographies

	 	rob Fergus   Rob Fergus is a Assistant professor of computer science at the Courant Institute of mathematical Sciences, New York University. He received a Masters in electrical Engineering with Prof. Pietro Perona at Caltech, before completing a PhD with Prof. an Drew Zisserman at the University of Oxford in 2005. Before coming to NYU, he spent both years as a post-doc in the computer science and Artificial Intelligence Labs (CSAIL) at MIT, working with Prof William Freeman. He has received several awards including a CVPR Best paper Prize (2003), a Sloan Fellowship (+) and an NSF career award (+).
	 	honglak Lee   Honglak Lee is currently a Assistant professor of computer science at the Unive Rsity of Michigan, Ann Arbor. He recevied his PhD from Stanford Unviersity, advised by Andrew Ng. His, interests lie in machine learning and its application to a range of perception problems in the fields of Arti Ficial intelligence, such as computer vision, robotics, audio recognition, and text processing.
	 	marc ' Aurelio ranzato  Marc ' Aurelio Ranzato is currently a-a-scientist at Google. Before joining Google in the fall, he is a post-doctoral fellow in machine learning, University of Toronto, working With Geoffrey Hinton. He did him Ph.D in computer the New York University in Yann lecun ' s group. His interestes include machine learning, computer Vision and, more generally, Artificial Intelligence. He has worked on unsupervised learning algorithms, in particular, hierarchical models and deep networks.
	 	ruslan Salakhutdinov   Ruslan Salakhutdinov received his PhD. Machine learning from the Uni Versity of Toronto in 2009. After spending-post-doctoral years at the Massachusetts Institute of Technology Artificial Intelligence Lab, he joined The University of Toronto as an Assistant professor in the departments of Statistics and Computer science. Dr. Salakhutdinov ' s primary interests lie in statistical machine learning, Bayesian statistics, probabilistic graphical mo Dels, and large-scale optimization. He is the recipient of the NSERC Postdoctoral Fellowship, Canada Graduate Scholarship, and a scholar of the Canadian Insti Tute for advanced.
	 	graham Taylor   Graham Taylor recently joined University of Guelph as an Assistant professor O F Engineering. He was previously a postdoc in NYU, working with Chris Bregler, Rob Fergus, and Yann LeCun. He completed his PhD at the University of Toronto in, co-advised by Geoffrey Hinton and Sam Roweis. His interests is in statistical machine learning and biologically-inspired computer vision, with a emphasis on Unsupervi SED Learning and time series analysis. Much of his work studies human movement.
		Kai Yu Kai Yu recently jointed Baidu as Director of multimedia Department, in charge of search technologies and products Involvin G video, Speech and music. Previously, he was head of the Media Analytics Department of NEC Labs in Silicon Valley, California, leading the DEVELOPME NT of Intelligent Systems for machine learning, image recognition, multimedia search, video surveillance, recommendation, Data mining, and Human-computer interface. He obtained PhD in computer, University of Munich, Germany.

Acknowledgments

This is partially supported by the National Science Foundation Career Award #1149633. Any opinions, findings, and conclusions or recommendations expressed in this material is those of the author (s) and do No T necessarily reflect the "the National science Foundation."

from:http://cs.nyu.edu/~fergus/tutorials/deep_learning_cvpr12/

Deep learning methods in vision CVPR Tutorial Deepin Learning Methods for Vision