An Active Learning Algorithm with Select Samples in Wide Area

摘要

In this paper, we propose an improved learning algorithm, called SWA (samples in wide area), base on with Gaussian kernel function, which is used to analyze the distribution of training set. To make the selection engine in active learning more inclined to choose new unlabeled samples from the area uncovered with training data samples. Such then, the training set will be scattered, to access higher classification accuracy. To evaluate the performance of the proposed algorithm, we conduct different experiments on four data sets. And experimental results show that SWA is superior to other comparison algorithms, including MS and EQB.