DISCERN: Diversity-based Selection of Centroids for k-Estimation and Rapid Non-stochastic Clustering

TL;DR

DISCERN is a deterministic method for estimating the number of clusters and initializing centroids in K-Means, improving clustering accuracy and efficiency compared to existing methods.

Contribution

The paper introduces DISCERN, a novel deterministic approach for estimating K and initializing centroids, enhancing K-Means clustering performance.

Findings

DISCERN outperforms other K estimation methods in accuracy.

Clustering results with DISCERN are more stable and accurate.

DISCERN improves both K estimation and final clustering quality.

Abstract

One of the applications of center-based clustering algorithms such as K-Means is partitioning data points into K clusters. In some examples, the feature space relates to the underlying problem we are trying to solve, and sometimes we can obtain a suitable feature space. Nevertheless, while K-Means is one of the most efficient offline clustering algorithms, it is not equipped to estimate the number of clusters, which is useful in some practical cases. Other practical methods which do are simply too complex, as they require at least one run of K-Means for each possible K. In order to address this issue, we propose a K-Means initialization similar to K-Means++, which would be able to estimate K based on the feature space while finding suitable initial centroids for K-Means in a deterministic manner. Then we compare the proposed method, DISCERN, with a few of the most practical K estimation methods, while also comparing clustering results of K-Means when initialized randomly, using K-Means++ and using DISCERN. The results show improvement in both the estimation and final clustering performance.