Data and Vision Weekly Seminar, CS, UTSA

Data and Vision Weekly Seminar

Department of Computer Science
University of Texas at San Antonio

Organizers: Qi Tian, Kay Robbins, Weining Zhang, and Yufei Huang (EE).
Time: 3-4 pm, Every Friday
Place: SB 3.01.02, CS Conference Room

Schedule for Fall 2004

10/01/04, Dimension Reduction Techniques: A Review
Speaker: Jerry Yu
Presentation slides
Abstract:
Advances in data collection and storage capacities lead to information overload in many fields. Traditional statistical methods often break down because of the increase in the number of variables in each observations, that is, the dimension of the data. Dimension reduction has been one of the most challenging problem in data exploring. In my presentation two traditional dimension reduction methods, Principal Component Analysis (PCA) and Multidimensional Scaling (MDS) will be reviewed and three relatively new ones, Isomap, Local Linear Embedding and Charting, will be introduced.
References:
10/08/04 CS Colloquium Series
10/15/04, Independent Component Analysis and Its Applications
Speaker: Qing Xue
Presentation slides
Abstract:
In the disciplines of statistics, data analysis, signal processing, and neural network research, a common problem is to find a suitable representation of multivariate data. A linear representation of the original data is often desirous for computational and conceptual simplicity. A recently developed linear transformation method is independent component analysis (ICA), in which the desired representation is the one that minimizes the statistical dependence of the components of the representation. What distinguishes ICA from other well-known linear transformation methods such as principal component analysis, factor analysis and projection pursuit is that it looks for components that are both statistically independent and nongaussian. In this talk we briefly introduce the basic concepts, applications and estimation principles of ICA.
References:
10/22/04, An Efficient Video Object Extraction Approach
Speaker: Like Zhang
Presentation slides
Abstract:
Video object extraction is important to video surveillance systems and MPEG4 implementations. Here we first have a brief review for commercial video surveillance products and researches on object extraction. Then we are going to talk about the drawbacks of previous pixel-based methods and propose an object-based backgroud updating approach, which incorporates object information to update background intelligently. Edge difference is adopted for object boundary extraction, and we are going to talk about detailed steps for extracting video objects. Experiment results and sample video clips are given to demonstrate the effectiveness of the algorithm.
Reference

C. Kim and J.-N. Hwang, “"Object-based video abstraction for video surveillance systems”" IEEE Transactions on Circuits and Systems for Video Technology, vol.12, no. 12, pp. 577-586, Dec. 2002. [PDF]
S. Chien, S. Ma, and L.-G. Chen, "“Efficient moving object segmentation algorithm using background registration Technique”" IEEE Transactions on circuits and systems video technology, vol.12, pp. 577-586, July 2002.[PDF]
R. Cucchiara, C. Grana, M. Piccardi, and A. Prati, "“Detecting moving objects, ghosts, and shadows in video streams”" IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, no. 10, pp. 1337-1341, Oct. 2003. [PDF]

10/29/04
No seminar.

11/05/04, Introduction to Image Super-resolution
Speaker: Kevin X. Su
Presentation slides
Abstract:
With the development of image processing applications, there is a big demand for high-resolution (HR) images since HR images not only give the viewer a pleasing picture but also offer additional detail that is important for the analysis in many applications. The current technology to obtain HR images mainly depends on sensor manufacturing technology that attempts to increase the number of pixels per unit area by reducing the pixel size. However, the cost for high-precision optics and sensors may be inappropriate for general purpose commercial applications, and there is a limitation to pixel size reduction due to shot noise encountered in the sensor itself. Therefore, a resolution enhancement approach using signal processing techniques to obtain an HR image (or sequence) from observed single/multiple low-resolution (LR) images has been a great concern in many areas, and it is called super-resolution (SR). In this presentation, a brief introduction to SR techniques will be given. Experimental results are used to demonstrate the effectiveness of the approaches.

References:

S.C. Park, M.K. Park, and M.G. KANG, "Super-Resolution Image Reconstruction: A Technical Overview", IEEE Signal Processing Magazine, Vol. 20, pp. 21-36, May 2003. [PDF]
W.T. Freeman, T.R. Jones, and E.C. Pasztor, "Example-Based Super-Resolution", IEEE Computer Graphics and Applications, Vol. 22, pp. 56-65, 2002. [PDF]

11/12/04, A Brief Introducation to Graphical Models
Speaker: Yijuan Lu
Presentation slides
Abstract:
Graphical models are a marriage between probability theory and graph theory. They provide a natural tool for dealing with two problems that occur throughout applied mathematics and engineering -- uncertainty and complexity-- and in particular they are playing an increasingly important role in the design and analysis of machine learning algorithms.
In my presentation, I will discuss the following topics: a) Definition: What are graphical models? b) Representation: How can a graphical model compactly represent a joint probability distribution? c) Inference: how can we infer the hidden state of a system? d) Learning: how do we estimate the parameters or structure of the model? e) Applications: what has this machinery been used for?

References:

K. Murphey, "Introduction to Graphical Models," Technical Report, May 2001. [PDF]
M. I. Jordan, Learning in Graphical Models, MIT Press, 1999.

11/19/04, A Bayesian Approach for Reconstructing Genetic Regulatory Networks with Hidden Factors
Speaker: Jiaying Wang
Presentation slides
Abstract

Based on large volumes of gene expression data obtained from micro-array experiments, it is an important yet difficulty task to identify (or reverse engineer) transcriptional networks of the underlying genetic system. In this presentation, I will discuss how graphical modeling and Bayesian methods can be used to infer the networks of human T-cell using the time series micro-array data. I will first introduce the modeling of the system with a special graphical model called the linear dynamic state space model (LDSSM). LDSSMs allow the existence of hidden factors and hidden factors are introduced to capture effects not directly measured in a micro-array experiment and thus are indispensable in accurate modeling the genetic systems. Next, I will discuss the parameter and structure inference of the LDSSM under a Bayesian framework using a Variational Bayes algorithm. In the end, I will discuss the testing procedure and demonstrate the inference results.
Reference

Beal,M.J., Falciani,F., Lioumi,M., Rangel,C. and Wild, D.L. (2004) A Bayesian approach to reconstructing genetic regulatory networks with hidden factors. Bioinformatics, Advance Access published online on September 7, 2004.
Beal, M. J. (2003). Variational Algorithms for Approximate Bayesian Inference. Ph. D. thesis, Gatsby Computational Neuroscience Unit, University College London, UK.
Rangel, C., J. Angus, Z. Ghahramani, M. Lioumi, E. Sotheran, A. Gaiba, Wild, D. L. and Falciani, F. (2004). Modelling T-cell activation using gene expression profiling and state space models. Bioinformatics 20, 1361-1372.

11/26/04
No seminar, Thanksgiving Days

Questions and Comments?

Please send emails to qitian@cs.utsa.edu, or seminar co-organizers: Kay Robbins, Weining Zhang, Yufei Huang, and Qi Tian.