GUC: UNSUPERVISED NON-PARAMETRIC GLOBAL CLUSTERING AND ANOMALY DETECTION
Keywords: Global Clustering, Distance-Based Clustering, Global Outlier Detection, Unsupervised Representation Learning
TL;DR: A novel algorithm for performing Global Clustering and Outlier Detection that achieves state-of-the-art performance
Abstract: Clustering is a crucial task in the fields of Machine and Representation Learning, with the ability of grouping similar data points together being of particular importance. In this paper, we propose a novel non-parametric algorithm that performs Global Clustering and Anomaly Detection in an unsupervised manner. Our algorithm is both effective and efficient, requiring no prior assumptions or domain knowledge to be applied. It features two modes that utilize the distance from the dataset’s center for clustering data points together. The first mode splits the dataset into global clusters where each cluster signifies proximity from the center. The second mode employs a threshold value for splitting the points into outliers and inliers.
We evaluate our proposal against other prominent methods using synthetic and real datasets. Our experiments demonstrate that the proposed algorithm achieves state-of-the-art performance with minimum computational cost, and can successfully
be applied to a wide range of Machine Learning applications.
Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 123
Loading