Choosing the Most Effective Pattern Classification Model under Learning-Time Constraint
Arquivos
Data
2015-06-26
Autores
Saito, Priscila T. M.
Nakamura, Rodrigo Y. M.
Amorim, Willian P.
Papa, Joao P. [UNESP]
Rezende, Pedro J. de
Falcao, Alexandre X.
Título da Revista
ISSN da Revista
Título de Volume
Editor
Public Library Science
Resumo
Nowadays, large datasets are common and demand faster and more effective pattern analysis techniques. However, methodologies to compare classifiers usually do not take into account the learning-time constraints required by applications. This work presents a methodology to compare classifiers with respect to their ability to learn from classification errors on a large learning set, within a given time limit. Faster techniques may acquire more training samples, but only when they are more effective will they achieve higher performance on unseen testing sets. We demonstrate this result using several techniques, multiple datasets, and typical learning-time limits required by applications.
Descrição
Palavras-chave
Como citar
Plos One. San Francisco: Public Library Science, v. 10, n. 6, 23 p., 2015.