December 2012′s outstanding undergrad research on SIURO

SIAM Undergraduate Research Online, SIAM’s web-based publication often referred to as SIURO, publishes outstanding undergraduate research in applied and computational mathematics. Two types of articles can be found in SIURO: papers to which undergraduates have made a significant contribution, and expository (survey) papers of high quality written by a faculty member or researcher for an undergrad audience.

The latest three papers published last month in SIURO provide insights to cluster analysis in large data sets, approximation methods for spectral clustering, and a review of numerical computational packages comparable to Matlab.

An Approach To Identify the Number of Clusters by Katelyn Gao (Massachusetts Institute of Technology), Heather Hardeman (University of Montevallo), Edward Lim (Johns Hopkins University), and Cristian Potter (East Carolina University), addresses statistical methods used to interpret and analyze big data in this technical age in which we live. The paper proposes a complementary method to one that uses eigenvalues to compute the number of clusters for finding patterns in data.

Study of Free Alternative Numerical Computation Packages by Matthew Brewster (University of Maryland, Baltimore County) investigates viable alternatives to Matlab for uses in teaching and research. Matlab is the most popular commercial package for numerical computations in mathematics, statistics, and other fields, but other free packages include Octave, Free Mat and Scilab, which possess many of the same features as Matlab.

Spectral Clustering: An empirical study of Approximation Algorithms and its Application to the Attrition Problem by A. Thompson, B. Cung, T. Jin, and J. Ramirez (University of Nebraska – Lincoln), evaluates approximation methods that have been developed for spectral clustering, which can become computationally expensive for large datasets. Approximation methods can help reduce running time even while maintaining accurate classification. By analyzing the empirical performance of existing approximate spectral clustering methods, the study addresses an important issue in business optimization.

Additional papers will be posted online as they are accepted.

Print Friendly
Facebook Twitter Email

Comments are closed.