Friday, 28 July 2017: 2:45 PM
Constellation E (Hyatt Regency Baltimore)
Principal Component Analysis is generally used either to reduce the dimensionality of data (e.g., for compression, or feature extraction), or to provide an explanation/interpretation of the underlying structure of data. Both of these goals, however, become increasingly problematic as the number of variables increases or exceeds the number of cases. In such situations, it has been proposed that principal components should additionally be constrained to be sparse (i.e., to have zero loadings on many/most of the variables). In this talk, the construction of the Sparse Principal Component Analysis is reviewed, and the method is applied to a number of example data sets from meteorology.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner