Thesis topic proposal
Károly Héberger
Fair comparison of similarity and distance measures (metrics)


Institute: University of Pécs
Doctoral School of Chemistry

Thesis supervisor: Károly Héberger
Location of studies: MTA TTK
Abbreviation of location of studies: MTA

Description of the research topic:

One of the most important topics in modern chemometric research is the fair comparison of existing methods and algorithms. Some of the techniques do suit better to special applications others provide average (not excellent) performances, but always and generally. To select the best analytical technique, model for a given purpose is a straightforward aim of our research project. Similarity measures (correlation coefficients: Pearson, Kendall, Spearman, concordant) and on a reversed scale dissimilarity measures (distances: Euclidean, Manhattan, Mahalanobis, etc.) are to be used, but the results are generally different even in case of proper data preprocessing (standardization, normalization range scaling, etc.). The procedure based on sum of (absolute) ranking differences developed in our laboratory is suitable for a fair comparison of methods, models features. It can be established, which similarity measure should be used and under which circumstances. Different data sets will be used for validation of novel methods and demonstrate the applicability and usefulness of novel techniques.

Deadline for application: 2017-01-31

All rights reserved © 2007, Hungarian Doctoral Council. Doctoral Council registration number at commissioner for data protection: 02003/0001. Program version: 1.2318 ( 2016. XI. 26. )