Clustering Using Conditional Generative Adversarial Networks (cGANs)
Marek Ružička, Matúš Dopiriak · 2023
Clustering is widely acknowledged as an essential process for accomplishing various tasks, such as document clustering, social news clustering, and sentiment analysis. However, conventional clustering algorithms often require the prior determination of the number of clusters to produce satisfactory results. In this study, we propose a novel clustering method that does not necessitate the number of clusters as a compulsory input parameter. Instead, we utilize a conditional Generative Adversarial Network (cGAN) to perform the clustering task. Our experimental results demonstrate that the proposed approach strikes a balance between processing time and accuracy, outperforming conventional clustering methods such as K-Means and Mean-Shift. These methods require the number of clusters to be specified beforehand or demand a significant amount of time to successfully complete the task. In contrast, our proposed method can efficiently cluster data without prior knowledge of the number of clusters, yielding promising outcomes.