A Stacked Denoising Autoencoder Based on Supervised Pre-training

Deep learning has attracted much attention because of its ability to extract complex features automatically. Unsupervised pre-training plays an important role in the process of deep learning, but the monitoring information provided by the sample of labeling is still very important for feature extraction. When the regression forecasting problem with a small amount of data is processed, the advantage of unsupervised learning is not obvious. In this paper, the pre-training phase of the stacked denoising autoencoder was changed from unsupervised learning to supervised learning, which can improve the accuracy of the small sample prediction problem. Through experiments on UCI regression datasets, the results show that the improved stacked denoising autoencoder is better than the traditional stacked denoising autoencoder.