Variational Autoencoder Model Combining Deep Learning and Probability Statistics: Research And Application

Abstract

A multi-layer generative model is proposed to improve the accuracy of large-scale data analysis by addressing the problems of insufficient feature extraction ability and insufficient association with label information in existing topic models. The model is divided into three main modules: text encoding, self-coding inference, and layered learning. The Poisson gamma confidence network is used as the decoder, and random gradient Monte Carlo sampling is employed as the posterior inference model. By utilizing the Fisher information matrix, the learning rates of different levels and topic parameters are adaptively adjusted. The layered learning strategy is introduced to construct a learning network, and based on this, text data and label information are combined for feature extraction. The results showed that the testing error rates of the research model on the 20News, RCV1, and IMDB datasets were 16.52%, 18.72%, and 11.67%, respectively, all of which were the lowest, and the testing time was the shortest. Therefore, research models have high data analysis and interpretation capabilities, and relatively high computational efficiency, which can provide scientific tools for accurate analysis of large-scale data in batches.