Extensions of principal components analysis
نویسنده
چکیده
To my father iii ACKNOWLEDGEMENTS My journey through the PhD program has been a long one, taking me from robotics, to computer vision, to machine learning, and finally to theory. Making so many transitions slowed my graduation, no doubt, but it also allowed me to work with leaders across several research fields and has given me a better perspective on the The most important man in my PhD career has been my advisor, Santosh Vem-pala, who brought out the best in me. Santosh's deep knowledge of theory and insight into my character allowed him to steer me toward the problems where I would enjoy myself and have success. Working with him closely has taught me how to think about math more intuitively and shown me how a truly great mind thinks. I will never forget his many acts of kindness and sometimes heroic efforts on my behalf. If I can emulate his advisement with my own students, I will consider myself a great success. I also want to thank my two previous faculty advisors Frank Dellaert and Jim Rehg for the opportunity to work with them and for all that I learned under their guidance. Matt Mullin has also been a great resource on all matters mathematical. I am also obliged to some of the great teachers at Georgia Tech. A few that stand out in my memory are committee also deserves thanks for their feedback and thoughtful questions about my work. Among my fellow graduate students, I am grateful to the members of the Theory Group and Wall Lab for their comradery through our shared struggles. I especially enjoyed my conversations with Jie Sun about our efforts at self-improvement and with iv Raffay Hamid about deep questions outside of computer science. These conversations were often the highlight of my day. Lastly, I want to thank my family. My parents more than anyone else cultivated the intellectual curiosity which has enriched my life and been a consolation through hard times. They and my wife Stephanie remained supportive throughout my PhD career. Stephanie has also been an excellent proofreader for all my papers. She and our two children, Charles and Sophie, have been remarkably patient through the late nights I spent in the lab and my bad humors. They mean the world to me. 2 Mapping points to the unit circle and then finding the direction of maximum variance reveals the …
منابع مشابه
A new weighting approach to Non-Parametric composite indices compared with principal components analysis
Introduction of Human Development Index (HDI) by UNDP in early 1990 followed a surge in use of non-parametric and parametric indices for measurement and comparison of countries performance in development, globalization, competition, well-being and etc. The HDI is a composite index of three indicators. Its components are to reflect three major dimensions of human development: longevity, knowledg...
متن کاملPersian Handwriting Analysis Using Functional Principal Components
Principal components analysis is a well-known statistical method in dealing with large dependent data sets. It is also used in functional data for both purposes of data reduction as well as variation representation. On the other hand "handwriting" is one of the objects, studied in various statistical fields like pattern recognition and shape analysis. Considering time as the argument,...
متن کاملEvaluation and Geographical analysis of the principal components affecting urban economic sustainability, Case study: Cities of Chaharmahal and Bakhtiari Province
Abstract Aims & Backgrounds: Today, economic challenges are one of the most important obstacles to achieving sustainability in the cities of developing countries. Therefore, recognition and geographical analysis of the factors affecting the economic sustainability of cities are among the important goals and priorities of urban and regional planning. Methodology: This research has been done by q...
متن کاملFunctional Analysis of Iranian Temperature and Precipitation by Using Functional Principal Components Analysis
Extended Abstract. When data are in the form of continuous functions, they may challenge classical methods of data analysis based on arguments in finite dimensional spaces, and therefore need theoretical justification. Infinite dimensionality of spaces that data belong to, leads to major statistical methodologies and new insights for analyzing them, which is called functional data analysis (FDA...
متن کاملGaussianizing Transformations for ICA
Nonlinear principal components analysis is shown to generate some of the most common criteria for solving the linear independent components analysis problem. These include minimum kurtosis, maximum likelihood and the contrast score functions. In this paper, a topology that can separate the independent sources from a linear mixture by specifically utilizing a Gaussianizing nonlinearity is demons...
متن کاملFisher Lecture: Dimension Reduction in Regression1,2
Beginning with a discussion of R. A. Fisher’s early written remarks that relate to dimension reduction, this article revisits principal components as a reductive method in regression, develops several model-based extensions and ends with descriptions of general approaches to model-based and model-free dimension reduction in regression. It is argued that the role for principal components and rel...
متن کامل