Predicting relative performance of classifiers from samples

• Type: Article in International Conference Proceedings Book
• Date: 2005

Authors

Leite, R (Author) Other; Pavel Brazdil (Author) FEP

Conference proceedings International

Title: ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning

Pages: 497-504

ICML 2005: 22nd International Conference on Machine Learning

Bonn, 7 August 2005 through 11 August 2005

Indexing

Scopus - 40 Citations

Other information

Authenticus ID: P-007-EY7

DOI: 10.1145/1102351.1102414

Abstract (EN): This paper is concerned with the problem of predicting relative performance of classification algorithms. It focusses on methods that use results on small samples and discusses the shortcomings of previous approaches. A new variant is proposed that exploits, as some previous approaches, meta-learning. The method requires that experiments be conducted on few samples. The information gathered is used to identify the nearest learning curve for which the sampling procedure was carried out fully. This in turn permits to generate a prediction regards the relative performance of algorithms. Experimental evaluation shows that the method competes well with previous approaches and provides quite good and practical solution to this problem.

Language: English

Type (Professor's evaluation): Scientific