Nearest Neighbor Searches on the GPU

Abstract

We introduce a GPU grid-based data structure for massively parallel nearest neighbor searches for dynamic point clouds. The implementation provides real-time performance and it is executed on GPU, both grid construction and nearest neighbors (approximate or exact) searches. This minimizes the memory transfer between device and system memories, improving overall performance. The proposed algorithm may be used across different applications with static and dynamic scenarios. Moreover, our data structure supports three-dimensional point clouds and given its dynamic nature, the user can change the data structure's parameters at runtime. The same applies to the number of neighbors to be found. Performance comparisons were made against previous works, endorsing the benefits of our solution. Finally, we were able to develop a real-time Point-Based Rendering application for validation of the data structure. Its drawbacks and data distribution's impact on performance were analysed and some directions for further investigation are given.

Accelerating adaptive inverse distance weighting interpolation algorithm on a graphics processing unit

“…Among these steps, the most computationally intensive step is to find the k nearest neighbours (kNN) for each interpolated point. Several effective kNN algorithms have been developed by region partitioning using various data structures [21,[44][45][46]. However, these algorithms are computationally complex in practice, and are not suitable to be used in implementing AIDW.…”

GPU-aided edge computing for processing the k nearest-neighbor query on SSD-resident data

“…According to this subdivision of the space, a GPU grid-based data structure is appropriate for massively parallel nearest neighbor searches over dynamic point datasets. A key contribution is [35] , where a grid-based indexing solution for 3-dimensional k -NN searches on the GPU is proposed. The k -NN algorithm works as follows: for a given query point, the algorithm expands the number of grid cells searched to ensure that at least k neighbors are found.…”

Fast ANN for High‐Quality Collaborative Filtering

“…To reduce the search space, the data can also be partitioned into a set of randomly overlapping spheres [Cay10]. If the structure of the data is known, such as 3D points, performance can be further improved [ZHWG08, QMN09, LSP*12, LTF*12]. Combining with accelerated radix sort [MG11], the process of sorting potential candidates can be speeded up significantly.…”