A fast k-means algorithm based on multi-granularity

Qing Wen; Junkuan Wang; Yabin Shao; Zizhong Chen

doi:10.1117/12.2628453

22 April 2022 A fast k-means algorithm based on multi-granularity

Qing Wen, Junkuan Wang, Yabin Shao, Zizhong Chen

Proceedings Volume 12174, International Conference on Internet of Things and Machine Learning (IoTML 2021); 1217410 (2022) https://doi.org/10.1117/12.2628453
Event: International Conference on Internet of Things and Machine Learning (IoTML 2021), 2021, Shanghai, China

Abstract

The k-means algorithm has been widely used since it was proposed, but the standard k-means algorithm does not perform well in terms of efficiency when dealing with large-scale data. To solve this problem, in this paper, we propose a fast kmeans algorithm based on multiple granularities. First, from the coarse-grained perspective, we use the clustering distribution information to narrow the search range of sample points, which makes the proposed algorithm very advantageous on large k. Second, from the fine-grained perspective, we use the rules of upper and lower bounds to reduce the number of sample points involved in the distance calculation, thus reducing many unnecessary distance calculations. Finally, we evaluate the proposed k-means algorithm on several real-world datasets, and the experimental results show that the proposed algorithm converges hundreds of times faster than standard k-means on average with the accuracy loss controlled at about three percent, and the speedup of the algorithm is more obvious when the dataset size is larger and the dimensionality of the dataset is higher.

Citation Download Citation

Qing Wen, Junkuan Wang, Yabin Shao, and Zizhong Chen "A fast k-means algorithm based on multi-granularity", Proc. SPIE 12174, International Conference on Internet of Things and Machine Learning (IoTML 2021), 1217410 (22 April 2022); https://doi.org/10.1117/12.2628453

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available