Constructive Learning Method Based on Human Semi-Supervised Algorithm
Li Pin · Journal of Xi'an University of Arts and Science · 2014
The application of Constructive Machine Learning( CML) algorithm training classification network needs large numbers of labeled examples which are hard to be retrieved. To address the issue,a constructive learning method based on human performance semi-supervised algorithm( HPSS) is proposed. According to the labeled examples,an initial classification network is constructed. Unlabeled examples can be selectively labeled by using the initial classification network to be incorporated into the labeled examples and the parameters of the classification network can be rectified. The process can be repeated until no newly-labeled examples are to be found. The final classification network is thus constructed. In the testing phase,the unlabeled samples are used again to label the testing examples in case of being indicated as denied. As the last step,the experiment is conducted on UCI data set. Results show that thesuggested algorithm is more effective than CML and Tri-CML algorithms.