Combination of multiple classifiers using probabilistic method
Heesung Lee, Sungjun Hong, Euntai Kim · 2007
The single neural network shows powerful classification ability. However, even increasing the size and number of hidden layers of the single network does not lead to improvements. In this paper, we propose the efficient multiple classifier combine method. We define the belief to represent the posterior probability of the pattern conditioned on all components of the classifiers. Since the probabilistic approach is the most promising tools in handling the uncertainty, proposed method can aggregate the results from the each neural network component efficiently. Experiments are performed with UCI machine learning repository to show the performance of the proposed algorithm.