An Effective Partitional Crisp Clustering Method Using Gradient Descent Approach

Enhancing the effectiveness of clustering methods has always been of great interest. Therefore, inspired by the success story of the gradient descent approach in supervised learning in the current research, we proposed an effective clustering method using the gradient descent approach. As a supplementary device for further improvements, we implemented our proposed method using an automatic differentiation library to facilitate the users in applying any differentiable distance functions. We empirically validated and compared the performance of our proposed method with four popular and effective clustering methods from the literature on 11 real-world and 720 synthetic datasets. Our experiments proved that our proposed method is valid, and in the majority of the cases, it is more effective than the competitors.