Depth variational auto-encoder model training method and device, equipment and storage medium

Abstract

The invention relates to a depth variational auto-encoder model training method and device, computer equipment and a storage medium. According to the invention, a pre-constructed depth variational self-encoding model can be trained by utilizing a visual chart training sample, the model can automatically learn valuable feature factors in the visual chart training sample without marking the chart in advance, and the generalization ability of the model is improved. The method comprises the steps of obtaining a visual chart training sample and a pre-constructed depth variational auto-encoder model framework; determining the dimension of the latent variable space and factor prior distribution of the latent variable space; and based on the dimension and factor prior distribution, training the pre-constructed depth variational auto-encoder model framework by using the visual chart training sample so as to enable the probability distribution of a reconstruction chart generated by the depth variational auto-encoder model framework to tend to be consistent with the original probability distribution of the visual chart training sample, and obtaining a trained depth variational auto-encoder model.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, in particular to a deep variational autoencoder model training method, device, computer equipment and storage medium. Background technique [0002] With the development of information technology, visualization software creation technology has gradually matured. Software such as Tableuau and Microsoft PowerBI greatly simplify the creation of visualization charts, and can provide intuitive and data-rich visualization charts for news articles, business reports, research papers, etc. [0003] Machine learning technology can identify valuable information from a large number of charts, such as chart patterns, including factors such as color, shape, position, etc. Based on these factors, chart features can be efficiently explored. For example, the deep variational autoencoder model (Deep Variational AutoEncoder, VAE) currently used can map a large number of complex graphs into ...

AI Technical Summary

Problems solved by technology

[0004] However, when training the deep variational autoencoder model, a large amount of sample data is often required, and for graph databases, such as heat map data of cities, the amount of data is small, which will lead to underfitting of the trained model and reduce the Generalization ability of deep variational autoencoder models, it is difficult to explore effective features

Images