DISCERN: diversity-based selection of centroids for k-estimation and rapid non-stochastic clustering

Abstract

As one of the most ubiquitously applied unsupervised learning methods, clustering has also been known to have a few disadvantages. More specifically, parameters such as the number of clusters and neighborhood radius are usually unknown and hard to estimate in practical cases. Moreover, the stochastic nature of a great number of these algorithms is also a considerable point of weakness. In order to address these issues, we propose DISCERN which can serve as an initialization algorithm for K-Means, finding suitable centroids that increase the performance of K-Means. Following that, the algorithm can estimate the number of clusters if need be. The algorithm does all of that, while maintaining complete robustness and returning the same results at each separate run. We ran experiments on the proposed method processing multiple datasets and the results show its undeniable superiority in terms of results, computational time and robustness when compared to the randomized K-Means and K-Means++ initialization. In addition, the superiority in estimating the number of clusters is also discussed and we prove the lower complexity when compared to methods such as the elbow and silhouette methods in estimating the number of clusters.

Tensor Graph Attention Network for Knowledge Reasoning in Internet of Things

“…Contrastive learning [31] is a discriminative paradigm in self-supervised learning [44] that allows the utilization of selfdefined pseudo labels as supervision to learn representations that are useful for various downstream tasks. It aims to encourage similar samples to cluster together while pushing diverse samples apart from each other.…”

“…To achieve in-depth mining of underlying logic, we use CompGCN [21] to update structural representations and employ self-attention [30] to iteratively capture local temporal features, which can ensure historical decay while avoiding the issue of feature stagnation caused by temporal sparsity. By incorporating localglobal features into contrastive learning [31], we generate a predictive mask vector that helps to narrow down the prediction scope and thus improves prediction accuracy.…”