English

Optimizing $k$ in $k$NN Graphs with Graph Learning Perspective

Machine Learning 2024-01-17 v1 Signal Processing

Abstract

In this paper, we propose a method, based on graph signal processing, to optimize the choice of kk in kk-nearest neighbor graphs (kkNNGs). kkNN is one of the most popular approaches and is widely used in machine learning and signal processing. The parameter kk represents the number of neighbors that are connected to the target node; however, its appropriate selection is still a challenging problem. Therefore, most kkNNGs use ad hoc selection methods for kk. In the proposed method, we assume that a different kk can be chosen for each node. We formulate a discrete optimization problem to seek the best kk with a constraint on the sum of distances of the connected nodes. The optimal kk values are efficiently obtained without solving a complex optimization. Furthermore, we reveal that the proposed method is closely related to existing graph learning methods. In experiments on real datasets, we demonstrate that the kkNNGs obtained with our method are sparse and can determine an appropriate variable number of edges per node. We validate the effectiveness of the proposed method for point cloud denoising, comparing our denoising performance with achievable graph construction methods that can be scaled to typical point cloud sizes (e.g., thousands of nodes).

Keywords

Cite

@article{arxiv.2401.08245,
  title  = {Optimizing $k$ in $k$NN Graphs with Graph Learning Perspective},
  author = {Asuka Tamaru and Junya Hara and Hiroshi Higashi and Yuichi Tanaka and Antonio Ortega},
  journal= {arXiv preprint arXiv:2401.08245},
  year   = {2024}
}
R2 v1 2026-06-28T14:17:52.058Z