International Conference on Computer and Knowledge Engineering

Home / 13th International Conference on Computer and Knowledge Engineering

Information Theoretic Learning-based Deep Embedded Clustering (ITL-DEC)

Authors :

Hoda Shad¹ Mona Zamiri² Tahereh Bahreini³ Reza Monsefi⁴ Ghoshe Abed Hodtani⁵

1- Computer Engineering Department Faculty of Engineering, Ferdowsi University of Mashhad (FUM), Azadi Square, Mashad, Iran 2- Computer Engineering Department Faculty of Wayne State University 3- Department of Electrical Engineering Ferdowsi University of Mashhad Mashhad, Iran 4- Computer Engineering Department Faculty of Engineering, Ferdowsi University of Mashhad (FUM), Azadi Square, Mashad, Iran 5- Department of Electrical Engineering Ferdowsi University of Mashhad Mashhad, Iran

Keywords :

Clustering،Deep Neural Networks،Autoencoder،Representation Learning،Unsupervised Learning،Cauchy-Schwarz Divergence،Jenson-shanon Divergence،Deep Clustering

Abstract :

Clustering as the best grouping algorithm for the data sets is a fundamental problem in many data-driven scientific and real-world applications. There are several methods based on some similarity measures for clustering, all suffering from high computational complexity on large-scale datasets. Clustering performance highly depends on the quality of data representation; hence, in the literature, various linear and nonlinear representation methods and deep learning-based clustering algorithms have been exploited. This paper presents a novel fully unsupervised deep clustering method with end-to-end training capable of simultaneously learning feature representations and cluster assignments using deep neural networks. We use autoencoder as our powerful feature extraction deep neural network and two information-theoretic divergence measures, Cauchy-Schwarz divergence and Jensen-Shannon divergence, as cost functions to train the network parameter and appropriate clustering feature space. Experiments performed on the benchmark data sets validate the effectiveness of the proposed method.