Easy Data Augmentation Method for Classification Tasks
Liu Guohang, Zhang Shibin, Haozhe Tang, Yang Lu, Jiazhong Lu, Huang Yuanyuan · 2020
In recent years, the rapid development of deep learning has made great progress in artificial intelligence-related technologies. With the deepening of artificial intelligence research, machine learning is applied to more and more fields. Although machine learning has many advantages and has achieved considerable results, machine learning and its related learning algorithms still face some related challenges. That is, the lack of sufficient training data or uneven class balance in the data set. In deep artificial neural networks, a large amount of training data is needed to learn effectively, and collecting such training data is often expensive and laborious. Data Augmentation overcomes this problem by artificially expanding the training set and label retention conversion. In this article, we propose a method to expand the data set based on ball k-means. The experiments on the Four-class, Digits, Iris and Breast-cancer data sets prove the efficiency of the method-fast execution speed, low computational complexity, and the effectiveness of the method. In these experiments, the quality of the training model has been improved, but the learning time stays the same as when the enhancement method is unused.