A comparison of Adaboosted FLD-based classifier with several standard classification algorithms
by Zehra Hasanain (2003-03-0075)
Time, Date & Venue
10:45 A.M. Saturday,
26th November 2005
Room 427 CS Dept.
Abstract:
The purpose of this survey is to compare the
performance of a relatively new algorithm, AdaboostFLD, against the performance
of several standard pattern classification algorithms, including neural
networks, decision trees, support vector machines, as well as another Adaboost-based
algorithm that uses decision stumps as the weak classifier. Seven binary
two-dimensional datasets have been used in the experiments, out of which four
are separable and three are non-separable. Training error, testing error and
training times have been compared for the above-mentioned algorithms. The
results obtained seem to show a favorable trend towards AdaboostFLD. This survey
describes the experiments that have been conducted, and discusses in detail the
results obtained in the course of this study.