Implementation method of self-reconfiguration k-means clustering technology based on soc-fpga

Abstract

The present invention discloses a SoC-FPGA-based self-reconfiguration K-means clustering technology implementation method, which includes the following steps: S1: constructing a SoC-FPGA heterogeneous platform model that cooperates between an ARM host end and an FPGA device end; S2: The ARM host side builds the OpenCL host program, creates the kernel, and completes memory allocation and mapping; S3: the host program calls the kernel program on the FPGA device side, and transmits data to the FPGA device side; S4: The first OpenCL kernel program calculates the Euclidean distance in parallel and pipelined, Generate a distance matrix; S5: self-reconfiguration of the second OpenCL kernel program, filter out the smallest element in each row and record its corresponding centroid; S6: self-reconfiguration of the third OpenCL kernel program, realize the distance of all sample points in each centroid cluster Accumulation and quantity statistics work; S7: The host program calculates new centroid data; S8: The host program performs iterative judgment. The invention not only improves the execution speed of the K-means clustering algorithm and obtains higher energy efficiency, but also solves the problem of insufficient FPGA hardware resources through self-reconfiguration of the kernel.

Description

technical field [0001] The invention relates to the technical field of data mining, in particular to a method for realizing a SoC-FPGA-based self-reconfiguration K-means clustering technology. Background technique [0002] As the most commonly used clustering algorithm, the K-means algorithm has been widely used in the fields of pattern recognition, machine learning and data mining due to its simplicity and effectiveness, such as automatic document classification, neural network basis function center determination, nuclear magnetic Resonance image segmentation processing, etc. K-means clustering belongs to unsupervised learning, and its process is as follows figure 1 As shown, the basic idea is based on the K centroids in the space, and the samples are classified according to the distance between the sample point and each centroid, and the new centroids of the current categories are calculated. The cluster centroids are updated multiple times until the centroids converge. ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art, provide a kind of execution speed that has promoted K-means clustering algorithm, obtain the self-reconfiguration K-means clustering technology realization based on SoC-FPGA of higher energy efficiency The method solves the problems of the existing K-means clustering algorithm in the prior art, such as large amount of computation, large hardware resource occupation, large power consumption, and large system delay.

Images