Village-Net clustering: A novel unsupervised manifold clustering method
Primary Area: learning on graphs and other geometries & topologies
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.
Keywords: Unsupervised clustering, Machine Learning, Random-Walks, Community detection
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: We present "Village-Net Clustering," a novel unsupervised clustering algorithm designed for effectively clustering complex manifold data. The algorithm operates in two primary phases: first, utilizing K-Means clustering, it divides the dataset into distinct "villages." Subsequently, a weighted network is created, where each node represents a village, capturing their proximity relationships. To attain the optimal clustering, we cluster this network using the Walk-likelihood Community Finder (WLCF), a community detection algorithm developed by one of our team members. An important feature of Village-Net Clustering is its ability to autonomously determine the optimal number of cluster. Extensive benchmarking on real datasets with known ground-truth labels showcases its competitive performance, particularly in terms of the normalized mutual information (NMI) score, when compared to state-of-the-art methods. Additionally, the algorithm demonstrates impressive computational efficiency, boasting a time complexity of O(N*k*d), where N signifies the number of instances, k represents the number of villages and d represents the dimension of the dataset, making it well-suited for effectively handling large-scale datasets.
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: 8865
Loading