Date: Tuesday May 12, 2009
Place: Room E305
Time: 4:00 PM
Speaker: Ritwik Kumar
Topic:Multi-fiber Reconstruction from DW-MRI data using a Continuous Mixture of Hyperspherical von Mises-Fisher Distributions
Paper Link: Here

Date: Thursday Mar 26, 2009
Place: Room E440
Time: 12:00 Noon
Speaker: Pavan Turaga
Topic:A System-of-Systems approach for summarizing Video Collections
Link: Here

Date: Thursday Mar 19, 2009
Place: Room E305
Time: 12:00 Noon
Speaker: Mohsen Ali
Topic:Nonrigid Structure From Motion in Trajectory Space
Paper Link: Here

Date: Friday Feb 26, 2009
Place: Room E440
Time: 12:00 Noon
Speaker: Jason Yu-Tseh Chi
Topic: Poisson Image Editing
Paper Link: Click Here

Date: Wednesday Feb 11, 2009
Place: Room 0032, Anderson Hall
Time: 3:00 pm
Speaker: Dr.Franjo Pernus
Topic: 3D/2D Registration for Image Guided Medical Interventions


Abstract:
Click here

Date: Friday Jan 30, 2009
Place: Room E404
Time: 2:50 pm
Speaker: Shun-Ichi Amari
Topic: Divergence measures between probability distributions and related information geometry.


Abstract:
TBA

Date: Friday Dec 5, 2008
Place: CSE 404 Room E122 (NOTE THE CHANGED VENUE)
Time: 12:00 pm
Speaker: Trevor Park
Topic: Some Methodology for Principal Component Regularization


Abstract:
Principal component analysis is widely used for many purposes, including dimension reduction and multivariate data exploration. Considerations of convenience, interpretation, and statistical stability often lead practitioners to modify the raw principal components to a more regular form, for example, one that has sparse loadings vectors. This talk will introduce one particular class of regularization methods based on penalized likelihood and discuss a computational challenge they pose: optimization on the orthogonal manifold.

Date: Friday Nov 21, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Venkatakrishnan Ramaswamy
Topic: Will be posted shortly.


Abstract:
Will be posted shortly.

Date: Friday Nov 14, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Dr. Pollina Golland
Topic: Will be posted shortly.


Abstract:
Will be posted shortly.

Date: Friday Nov 7, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Dr. Alper Ungor
Topic: Voronoi Diagrams and their Applications.


Abstract:
Voronoi diagram is a special kind of decomposition of a space determined by distances to a specified discrete set of objects in the space. There are many different types depending on the distance measure and type of objects. Voronoi diagrams find use in a variety of application fields, such as grahics, data mining, networks, machine learning and computer vision. This talk will review the basic definitions, results, as well as recent progress on the application of Voronoi diagrams.

Date: Friday Oct 31st, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Guang Cheng
Topic: Non-rigid registration

Date: Friday Oct 10, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: AmitDhurandhar
Topic: Semi-analytical Method for Analyzing Models and Model Selection Measures based on Moment Analysis

LINKS TO PAPERS:

Semi-analytical Method for Analyzing Models and Model Selection Measures based on MomentAnalysis. (Accepted to ACM Transactions on Knowledge Discovery and Data Mining - ACM TKDD)
Probabilistic Characterization of Random Decision Trees. (Accepted to Journal of Machine Learning Research - JMLR)
Probabilistic Characterization of Nearest Neighbor Classifier.

Abstract:
In this paper we propose a moment based method for studying models and model selection measures. By focusing on the probabilistic space of classifiers induced by the classification algorithm rather than on of datasets, we obtain efficient characterizations for computing the moments. By assuming the data to be drawn independently and identically from the underlying probability distribution and by going over the space of all possible datasets, we establish general relationships between the Generalization error, Hold-out-set error, Cross-validation error and Leave-one-out error. We later exemplify the method and the results by studying the behavior of the errors for the Naive Bayes Classifier. We also suggest ways of extending the analysis to other classification algorithms.

Date: Friday Oct 3, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Raazia Mazhar
Topic: EK-SVD: Optimized Dictionary Design for Sparse Representations.

LINK TO PAPER

Abstract:
Sparse representations using overcomplete dictionaries are used in a variety of field such as pattern recognition and compression. However, the size of dictionary is usually a tradeoff between approximation speed and accuracy. In this paper we propose a novel technique called the Enhanced K-SVD algorithm (EK-SVD), which finds a dictionary of optimized sizefor a given dataset, without compromising its approximation accuracy. EK-SVD improves the K-SVD dictionary learning algorithm by introducing an optimized dictionary size discovery feature to K-SVD. Optimizing strict sparsity and MSE constraints, it starts with a large number of dictionary elements and gradually prunes the under-utilized or similar-looking elements to produce a well-trained dictionary that has no redundant elements. Experimental results show the optimized dictionaries learned using EK-SVD give the same accuracy as dictionaries learned using the K-SVD algorithm while substantially reducing the dictionary size by 60%.

Date: Friday Sep 26, 2008
Place: CSE 404
Time: 12:00 pm
Speaker: Karthik Gurumoorthy
Topic: Sparse Projections Onto Exemplar Orthonormal Bases for Compact Image Representation.

LINK TO PAPER

Abstract:
We present a new method for compact representation of large image datasets. Our method is based on treating small patches from an image as matrices as opposed to the conventional vectorial representation, and encoding those patches as sparse projections onto a set of exemplar orthonormal bases, which are learned a priori from a training set. The end result is a low-error, highly compact image/patch representation that has significant theoretical merits and compares favorably with existing techniques on experiments involving the compression of ORL and Yale face databases.

Date: Wednesay Sep 17, 2008
Place: CSE E221
Time: 1:55 pm
Speaker: Dr.Amit Roy Chowdhury
Topic: From Single Images To Camera Networks: Modeling and Inference Strategies

Abstract:
The complexity of vision systems can be represented along many parameters, one of them being the amount of data that is processed. On one end of this spectrum is a single image, while on the other end is a large camera network. In this talk, he will focus on these two ends of the spectrum, analyze their unique requirements and inter-relationships.

In the first part, he will discuss mathematical models of image appearance. In my research, he tried to address the question on how valid are some of the commonly used models, like linear, bilinear, multilinear, locally linear. Given the physical laws of object motion, surface properties and image formation, can we derive some of these models from first principles? We will see that, under certain mathematical assumptions, we can indeed derive some of these models and that this analysis provides new insights into problems of tracking and recognition.

In the second part of the talk, I will discuss his current work on scene analysis in camera networks. He will first describe a multi-objective optimization framework that is able to hold tracks of multiple targets over space and time by adapting between delay and accuracy requirements. Then, he will describe his recent work on cooperative control of a camera network using game theory. The importance of a good understanding of the properties of single images in analyzing data over a camera network will be highlighted.