Infinite Ensemble for Image Clustering
Hongfu Liu*, Northeastern University; Ming Shao, Northeastern University; Sheng Li, Northeastern University; Yun Fu, Northeastern University
Image clustering has been a critical preprocessing step for vision tasks, e.g., visual concept discovery, content-based image retrieval. Conventional image clustering methods use handcraft visual descriptors as basic features via K-means, or build the graph within spectral clustering. Recently, representation learning with deep structure shows appealing performance in unsupervised feature pre-treatment. However, few studies have discussed how to deploy deep representation learning to image clustering problems, especially the uniﬁed framework which integrates both representation learning and ensemble clustering for eﬃcient image clustering still remains void. In addition, even though it is widely recognized that with the increasing number of basic partitions, ensemble clustering gets better performance and lower variances, the best number of basic partitions for a given data set is a pending problem. In light of this, we propose the Inﬁnite Ensemble Clustering (IEC), which incorporates the power of deep representation and ensemble clustering in a one-step framework to fuse inﬁnite basic partitions. Generally speaking, a set of basic partitions is ﬁrstly generated from the image data. Then by converting the basic partitions to the 1-of-K codings, we link the marginalized auto-encoder to the inﬁnite ensemble clustering with i.i.d. basic partitions, which can be approached by the closed-form solutions. Finally we follow the layer-wise training procedure and feed the concatenated deep features to K-means for ﬁnal clustering. Extensive experiments on diverse vision data sets with diﬀerent levels of visual descriptors demonstrate both the time eﬃciency and superior performance of IEC com-pared to the state-of-the-art ensemble clustering and deep clustering methods.