A k-nearest neighbor classification acceleration method based on opencl and soc-fpga

Abstract

The invention discloses a K-nearest neighbor classification acceleration method based on OpenCL and SoC-FPGA, which comprises the following steps: S1: building a SoC-FPGA heterogeneous platform model; S2: building an OpenCL host program for control at the ARM host end; S3 : The host program on the ARM host side creates the kernel; S4: The ARM host side configures and invokes the kernel program on the FPGA device side to transmit the data to the FPGA device side; S5: The first kernel program is responsible for calculating the Euclidean distance between the data to be classified and the training set data , to generate a distance matrix; S6: The second kernel program incompletely sorts each row of the distance matrix, screens out the minimum K distances in each row, and finds the corresponding training set element type, and sends it back to the ARM host for processing; S7: The ARM host performs category statistics and classification. The present invention uses the OpenCL standard to realize the FPGA parallel optimization of the K-nearest neighbor classification algorithm, and forms a system-level pipeline at the ARM end and the FPGA end. Compared with the traditional GPU heterogeneous computing system, the present invention has lower power consumption and higher efficiency. energy efficiency.

Description

technical field [0001] The invention relates to an acceleration method of K nearest neighbor classification technology based on SoC-FPGA novel heterogeneous computing system. Background technique [0002] As one of the top ten classic data mining algorithms in the 20th century, the K-nearest neighbor algorithm is widely used in text classification, pattern recognition, image and spatial classification and other fields due to its advantages of accuracy, simplicity and effectiveness. The K-nearest neighbor algorithm is based on lazy learning. Its basic idea is to find the K reference samples closest to each sample to be classified in the known training set, and determine the category of the sample to be classified according to the most category of the K reference samples. However, the K-nearest neighbor algorithm involves a lot of calculations, especially when the training set samples compared with the samples to be classified are large, it will bring a lot of calculation over...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art, provide a kind of acceleration method based on the K-nearest neighbor classification technology of SoC-FPGA novel heterogeneous computing system, solve the large amount of calculations and take up hardware resources that exist in the K-nearest neighbor algorithm of the prior art Multiple, high power consumption, and large system delays

Images