Binary tree of SVM: a new fast multiclass training and classification algorithm

• Published in: IEEE transactions on neural networks Vol. 17; no. 3; pp. 696 - 704
• Main Authors: Fei, B., Jinbai Liu
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.05.2006; Institute of Electrical and Electronics Engineers
• Subjects: Algorithms; Applied sciences; Artificial Intelligence; Binary tree of support vector machine (BTS); Binary trees; c-BTS; Classification; Classification algorithms; Classification tree analysis; Classifiers; Cluster Analysis; Complexity; Computer science; control theory; systems; Connectionism. Neural networks; Convergence; Error correction codes; Exact sciences and technology; Image Interpretation, Computer-Assisted - methods; Information Storage and Retrieval - methods; multiclass classification; Neural networks; Pattern Recognition, Automated - methods; probabilistic output; support vector machine (SVM); Support vector machine classification; Support vector machines; Testing; Training; Tree graphs; Trees; Upper bound; probabilistic output; Binary tree; multiclass classification; support vector machine (SVM); Classification; Vector support machine; Neural network; Binary tree of support vector machine (BTS); Acyclic graph; Multiclass; Directed graph; c-BTS
• DOI: 10.1109/TNN.2006.872343
• ISSN: 1045-9227

We present a new architecture named Binary Tree of support vector machine (SVM), or BTS, in order to achieve high classification efficiency for multiclass problems. BTS and its enhanced version, c-BTS, decrease the number of binary classifiers to the greatest extent without increasing the complexity of the original problem. In the training phase, BTS has N-1 binary classifiers in the best situation (N is the number of classes), while it has log 4/3 ((N+3)/4) binary tests on average when making a decision. At the same time the upper bound of convergence complexity is determined. The experiments in this paper indicate that maintaining comparable accuracy, BTS is much faster to be trained than other methods. Especially in classification, due to its Log complexity, it is much faster than directed acyclic graph SVM (DAGSVM) and ECOC in problems that have big class number