Fast global k-means clustering method accelerated using opencl

Abstract

The invention discloses a fast global K-means clustering method using OpenCL acceleration, the steps of realization are: (1) read in the data set and the total number of clusters; (2) transpose the data at the OpenCL hardware device end; (3) )Calculate the centroid as the first initial clustering center point; (4) cluster using the K-means algorithm; (5) select a new initial clustering center point on the OpenCL hardware device side; (6) judge the current clustering center Whether the total number of points is less than or equal to the total number of clusters, if so, execute step (4), if not, execute step (7); (7) output the clustering result. The invention can realize real-time processing of massive clustering data on any hardware device supporting the open computing language OpenCL.

Description

technical field [0001] The invention belongs to the technical field of data processing, and further relates to a Fastglobal K-means clustering method accelerated by OpenCL (Open Computing Language) hardware equipment in the technical field of data mining. The invention can realize the parallel accelerated fast global K-means clustering method, and can realize the real-time processing of massive data on the open computing language OpenCL hardware device. Background technique [0002] The fast global K-means algorithm uses a deterministic method that does not depend on any initial parameter value instead of random search to obtain the initial cluster center, which solves the problem of poor stability of the clustering results of the traditional K-means clustering method based on local search. And by optimizing the calculation method of global K-means to increase the new cluster center, the running time is reduced. [0003] TCL Group Co., Ltd. disclosed a patent document "A me...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that the parallelized K-means code implemented by the Spark platform cannot be transplanted on various hardware devices that support the open computing language OpenCL, and the portability is not high. It is only suitable for distributed systems and cannot Parallel acceleration on GPU, FPGA, MIC and other hardware devices greatly limits its application scope

However, this method still has the disadvantage that it needs to store a large amount of intermediate calculation results when accelerating the algorithm on the heterogeneous device side. When the input image data is too large, it exceeds the storage capacity of the heterogeneous device, resulting in This method cannot realize the processing of larger image data

Images