RI | Research | Projects | Component Analysis for Data Analysis

Text only version of this site

Component Analysis for Data Analysis Head: Fernando De la Torre Frade Contact: Fernando De la Torre Frade (ftorre@cs.cmu.edu) Mailing address: Carnegie Mellon University Robotics Institute 211 Smith Hall Pittsburgh, PA 15213 Associated center: VASC

Jump to: Project Description | Personnel | Publications

Project Description

Component Analysis methods (e.g. Kernel Principal Component Analysis, Independent Component Analysis, Tensor factorization) have been successfully applied to modeling, classification and clustering in numerous visual, graphics and signal processing tasks over the last four decades. CA techniques are especially appealing because many can be solved as generalized eigenvalue problems or alternated least squares procedures, for which there exist extremely efficiently and numerically stable algorithms. These spectral approaches offer a potential for solving linear and non-linear estimation/learning problems efficiently and without local minima. In this project, we develop a framework for energy-based learning of component analysis methods and apply it to improve state-of-the-art methods for classification (e.g. support vector machines), clustering (e.g. normalized cuts) or face tracking algorithms (e.g. active appearance models).

Personnel [Past Members]

	Name	Title	Email Address
	Fernando De la Torre Frade	Research Scientist	ftorre@cs.cmu.edu
	Takeo Kanade	U.A. and Helen Whitaker University Prof., RI/CS	tk@cs.cmu.edu

Recent publications [View all 11 publications]

• A Comparative Study of Supervised Learning Techniques for the Radiative Transfer Equation Inversion
  E. Garcia, F. De la Torre Frade, and A.D. Castro
  International Conference on Machine Learning and Data Analysis, October, 2007. [Abstract]

• Bilinear Active Appearance Models
  J. Gonzalez, F. De la Torre Frade, R. Murthi, N. Guil Mata, and E. Zapata
  Workshop on Non-rigid Registration and Tracking through Learning, October, 2007. [Abstract]
  Download: pdf [965 KB] copyrighted

• Indoor People Tracking based on Dynamic Weighted Multidimensional Scaling
  J. Maria Cabero, F. De la Torre Frade, I. Arizaga, and A. Sanchez
  IEEE International Symposium on Modeling, Analysis and Simulation of Wireless and Mobile Systems, October, 2007. [Abstract]
  Download: pdf [706 KB] copyrighted

• Filtered Component Analysis to Increase Robustness to Local Minima in Appearance Models
  F. De la Torre Frade, A. Collet Romea, J. Cohn, and T. Kanade
  IEEE Conference on Computer Vision and Pattern Recognition, June, 2007. [Abstract]
  Download: pdf [4071 KB] copyrighted

• Learning Kernel Expansions for Image Classification
  F. De la Torre Frade and O. Vinyals
  IEEE Conference on Computer Vision and Pattern Recognition, June, 2007. [Abstract]
  Download: pdf [830 KB] copyrighted

• Parameterized Kernels for Support Vector Machine Classification
  F. De la Torre Frade and O. Vinyals
  International Conference on Computer Vision Theory and Applications, March, 2007, pp. 207-213. [Abstract]

• Simultaneous registration and clustering for temporal segmentation of facial gestures from video
  F. De la Torre Frade, J. Campoy, J. Cohn, and T. Kanade
  2nd International Conference on Computer Vision Theory and Applications, March, 2007. [Abstract]

• Discriminative Cluster Analysis
  F. De la Torre Frade and T. Kanade
  International Conference on Machine Learning, ACM Press, New York, NY, USA, Vol. 148, June, 2006, pp. 241 - 248. [Abstract]
  Download: pdf [488 KB] copyrighted

• Multimodal Oriented Discriminant Analysis
  F. De la Torre Frade and T. Kanade
  International Conference on Machine Learning (ICML)., August, 2005. [Abstract]
  Download: pdf [401 KB] copyrighted

• Representational Oriented Component Analysis (ROCA) for Face Recognition with One Sample Image per Training Class
  F. De la Torre Frade, R. Gross, S. Baker, and V. Kumar
  Computer Vision and Pattern Recognition, Vol. 2, June, 2005, pp. 266 - 273. [Abstract]
  Download: pdf [559 KB] copyrighted