Convolutional Autoencoder-based Deep Embedded Fuzzy Clustering Using H-divergences
<p>Tianzhen Chen</p> · Academic Journal of Computing & Information Science · 2024
With the rapid development of deep learning technologies, the integration of fuzzy clustering methods with convolutional autoencoders has offered new perspectives for data clustering. This paper introduces a novel deep learning methodology, "Convolutional Autoencoder-based Deep Embedded Fuzzy Clustering Using H-divergences, (CADEFC)". In response to the rapid evolution of deep learning, this study combines fuzzy clustering methods with convolutional autoencoders to explore new dimensions in data clustering. We critically assess the limitations of the Wasserstein distance as a conventional loss function and propose the adoption of H-divergences as a more robust alternative. By integrating H-divergence and introducing fuzzy theory, our method transcends the traditional constraints of deep clustering techniques, offering substantial improvements in clustering accuracy and stability. The effectiveness and superiority of this approach are demonstrated through rigorous testing on several benchmark datasets, including Digits, Fashion-MNIST, MNIST, and USPS. Our results confirm that the proposed methodology not only enhances performance but also broadens the applicative landscape of deep embedded clustering.