In Partial Fulfillment of the Requirements for the Degree of Master of Science
will defend his thesis
An Effective Approach to External Cluster Assessment With an Application to WHAN Galaxy Classification
For the classification of galaxies, a lot of different properties like morphology, colors, spectral features, etc. can be used. Among all these available methods, Emission-line classification of galaxies are among the easiest to carry out and are capable of dealing with issues such as star formation, chemical composition or nuclear activity. However, a lot of surveys need to be performed by astronomers to derive the classification structure and also a couple of emission-line spectra which are the most important are used for classification. Clustering methods can reduce the effort of manual classification by deriving inherent structure of data using intrinsic properties. Cluster validation plays a key role in assessing the value of the output of a clustering algorithm by computing statistics over the clustering structure. Cluster validation can be of two types; internal, where the metrics calculated use clustered data objects only and external, where metrics are computed by comparing clustered data objects to an independent external classification scheme. A large separation between clusters and classes serves to indicate cluster novelty while, finding clusters resembling existing classes serves to confirm existing theories of data distributions. Our proposed method for external cluster assessment unlike traditional metrics which output a single value for the clustering algorithm, computes separation between each individual cluster and it’s most similar external class. We use a sample of dataset from the Sloan Digital Sky Survey Dataset (SDSS) to evaluate our algorithm.
Date: Tuesday, April 26, 2016
Time: 8:30 AM
Place: HBSB 350
Advisor: Dr. Ricardo Vilalta
Faculty, students, and the general public are invited.