基于信息传递的快速无参聚类

Cluster analysis is an important branch in the field of data mining, and its purpose is to divide a set of data into different clusters by mining the internal properties of these data. Existing graph clustering methods face the problem of high time complexity, making them difficult to handle large-scale datasets. Besides, most existing methods face the intractable hyper-parameter problem due to the introduction of various regularization terms. In addition, many graph clustering models require additional post-processing steps, resulting in the obtained solutions that are far from the discrete solutions obtained by solving the original problem directly. In order to solve the above problems, this paper proposes a fast parameter-free clustering model via information transmission. This model introduces the idea of balanced clustering, which not only can avoid the introduction of additional regularization terms, but also can utilize the prior knowledge of bipartite graphs to process large-scale datasets. In addition, the block coordinate descent method makes the model directly obtain the solution of the original discrete problem without post-processing steps. Experimental results on multiple datasets show that the proposed method has better performance than compared methods in most cases.