Combining Classifiers: From the Creation of Ensembles to the Decision Fusion
- 1 August 2011
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-10
Abstract
Multiple classifier combination methods can be considered some of the most robust and accurate learning approaches. The fields of multiple classifier systems and ensemble learning developed various procedures to train a set of learning machines and combine their outputs. Such methods have been successfully applied to a wide range of real problems, and are often, but not exclusively, used to improve the performance of unstable or weak classifiers. In this tutorial are presented the basic terminology of the field, a discussion on the effectiveness of combination algorithms, the diversity concept, methods for the creation of an ensemble of classifiers, approaches to combine the decisions of each classifier, the recent studies and also possible future directions.Keywords
This publication has 32 references indexed in Scilit:
- A study on the performances of dynamic classifier selection based on local accuracy estimationPattern Recognition, 2005
- Diversity creation methods: a survey and categorisationInformation Fusion, 2005
- Decision templates for multiple classifier fusion: an experimental comparisonPattern Recognition, 2001
- On the algorithmic implementation of stochastic discriminationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2000
- On combining classifiersIEEE Transactions on Pattern Analysis and Machine Intelligence, 1998
- The random subspace method for constructing decision forestsIEEE Transactions on Pattern Analysis and Machine Intelligence, 1998
- Error Correlation and Error Reduction in Ensemble ClassifiersConnection Science, 1996
- The Lack of A Priori Distinctions Between Learning AlgorithmsNeural Computation, 1996
- Neural Networks and the Bias/Variance DilemmaNeural Computation, 1992
- Adaptive Mixtures of Local ExpertsNeural Computation, 1991