Decision templates for multiple classi"er fusion: an experimental comparison
نویسندگان
چکیده
Multiple classi"er fusion may generate more accurate classi"cation than each of the constituent classi"ers. Fusion is often based on "xed combination rules like the product and average. Only under strict probabilistic conditions can these rules be justi"ed. We present here a simple rule for adapting the class combiner to the application. c decision templates (one per class) are estimated with the same training set that is used for the set of classi"ers. These templates are then matched to the decision pro"le of new incoming objects by some similarity measure. We compare 11 versions of our model with 14 other techniques for classi"er fusion on the Satimage and Phoneme datasets from the database ELENA. Our results show that decision templates based on integral type measures of similarity are superior to the other schemes on both data sets. ( 2000 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.
منابع مشابه
Combination of Classi ers on the Decision Level for Face Recognition
This report is divided into two parts. In the rst part we present an overview of sensorfusion. We analyze the single steps in a fusion process and describe in a systematic wayseveral methods for combining pattern classi ers. The second part describes some practicalexperiments with classi er combination in the eld of face recognition. We investigate theimpact of decision combinat...
متن کاملDecision templates for multiple classifier fusion: an experimental comparison
Multiple classifier fusion may generate more accurate classification than each of the constituent classifiers. Fusion is often based on fixed combination rules like the product and average. Only under strict probabilistic conditions can these rules be justified. We present here a simple rule for adapting the class combiner to the application. c decision templates (one per class) are estimated w...
متن کاملCombining Nearest Neighbor Classifiers Through Multiple Feature Subsets
Combining multiple classi ers is an e ective technique for improving accuracy. There are many general combining algorithms, such as Bagging or Error Correcting Output Coding, that signi cantly improve classi ers like decision trees, rule learners, or neural networks. Unfortunately, many combining methods do not improve the nearest neighbor classi er. In this paper, we present MFS, a combining a...
متن کاملNearest Neighbor Classi cation from Multiple Feature Subsets
Combining multiple classi ers is an e ective technique for improving accuracy. There are many general combining algorithms, such as Bagging, Boosting, or Error Correcting Output Coding, that signi cantly improve classi ers like decision trees, rule learners, or neural networks. Unfortunately, these combining methods do not improve the nearest neighbor classi er. In this paper, we present MFS, a...
متن کاملAudio-Visual Speaker Veri cation using Continuous Fused HMMs
This paper examines audio-visual speaker veri cation using a novel adaptation of fused hidden Markov models, in comparison to output fusion of individual classi ers in the audio and video modalities. A comparison of both hidden Markov model (HMM) and Gaussian mixture model (GMM) classi ers in both modalities under output fusion shows that the choice of audio classi er is more important than vid...
متن کامل