Multi-view enhanced image clustering method

Abstract

The invention provides a multi-view enhanced image clustering method, which belongs to the field of image clustering and enhanced learning, and comprises the following steps: 1) pre-training an independent feature extraction network of each view, and initializing a potential feature space of each view; 2) pre-training a multi-view feature fusion network, and initializing a fusion feature space of each view; 3) using a K-means method to initialize a clustering environment, and allocating Bernoulli units to clustering prototypes in the environment; 4) distributing random rewards in real time by using an online reward strategy, and dynamically updating Bernoulli units in the environment; and 5) updating parameters, iteratively optimizing the clustering prototype until a convergence condition is met, and completing a multi-view enhanced clustering process. According to the invention, the online reward strategy is combined with the learning fusion representation and adjustment cluster, the complementary information between the view angles and the interaction information between the sample and the clustering prototype are fully used in the clustering analysis process, and the clustering performance is effectively improved.

Description

technical field [0001] The invention belongs to the field of image clustering and reinforcement learning, and relates to a multi-view enhanced image clustering method. Background technique [0002] With the widespread application of technologies such as network information and e-commerce, human beings have more and more ways to obtain data and information, more and more data can be collected, more and more complex data structures, and higher data dimensions. Multi-view image data usually comes from different fields of data objects or measurement results from multiple angles, and contains rich complementary information, which can effectively enhance the effect of data analysis. Complementary information in perspective data is difficult to fully utilize. Therefore, it is urgent to study a new method to deeply mine the complementary information between massive multi-view image data. [0003] Clustering is an important data analysis and processing technology in the field of ma...

AI Technical Summary

Problems solved by technology

[0004] Although current multi-view clustering methods use deep generative models to capture complementary information between multiple views and achieve good clustering results, the existing multi-view clustering methods only consider the inherent properties of multi-view data and ignore It improves the correlation between multi-view data and cluster centers, and makes the edges of clusters gradually blurred.

Images