The cost of post-secondary education in the U.S. continues to grow faster than salaries and inflation. In fact, the real cost of a college education has climbed almost 30 in the past 10 years and shows no sign of stabilizing in the near future. The economic competitiveness of the country increasingly depends on a skilled workforce with a post-secondary education capable of dealing with the demands of the global market. Thus, college attainment is at the center of producing a skilled workforce, and so it is, post-secondary education affordability. Using the national post-secondary student aid surveyor the year 2003- 2004, which is representative of the entire undergraduate population in the U.S., this study examines the various ways students and families pay for post-secondary education and its subsequent effect on persistence and performance for all groups of students across racial/ethnic and social-economic status lines.We use a spectral clustering algorithm based on normalized cuts to classify students based on their similarities. More specifically, we construct a social network with the students as the nodes of the graph and edge between pair of the students is weighted based on their similarity in attributes. We then obtain three nontrivial smallest of the Laplacian matrix. We use these to perform a k-means clustering in the eigenspace. We were able to establish meaningful clusters by this approach that helps in classifying students based on the relation between their persistence level and conditions of living.

Social network analysis deals with the interactions between individuals by considering them as nodes of a network (graph) whereas their relations are mapped as network edges. Study of such structures lies on the intersection of two different areas of research: sociology and graph theory. In this paper we give an overview of the mathematical concepts used for studying these networks as well as the major methodologies employed for the study of them. Most prominent applications and dominant research trends of the field are also discussed.

This paper presents theoretical relationships among several generalized LDA algorithms and proposes computationally efficient approaches for them utilizing the relationships. Generalized LDA algorithms are extended nonlinearly by kernel methods resulting in nonlinear discriminant analysis. Performances and computational complexities of these linear and nonlinear discriminant analysis algorithms are compared.