.png){kind=icon}
Return
Multi-block projection clustering
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2026.116103.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Projection clustering provides an effective framework for high-dimensional data analysis by learning an explicit mapping from the original space to a low-dimensional representation. However, many existing methods construct neighborhood relationships over the entire dataset, resulting in high computational and memory costs that limit their scalability. To address this issue, we propose the multi-block projection clustering (MBPC), which first randomly partitions the samples into multiple blocks of approximately equal size, and then jointly learns block-wise adaptive neighborhoods, block-specific projections, and a shared projection. In this way, MBPC avoids full-scale similarity computation while preserving local geometric structures and integrating complementary information from different blocks into a unified projection. An efficient alternating optimization algorithm is developed to solve the proposed model. Extensive experiments on benchmark datasets show that MBPC achieves competitive clustering performance, while maintaining good scalability in terms of running time, memory consumption, and out-of-sample applicability.
Journal
IF:
8
Papers:
5.8K
Citations:
3.5W
Organization
Cited Papers
From few to many: illumination cone models for face recognition under variable lighting and pose
PREAI
ShareSaveShareSave
ShareSaveShareSaveClustering and projected clustering with adaptive neighbors.png)
Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining
Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data miningIF0
PREAI
ShareSave
View more