Mehmet Ali Altuncu, Bahadır Türkoğlu, Mehmet Ali Çavuşlu, Suhap Şahin
IEEE 26th Signal Processing and Communications Applications, 2018
K-means algorithm is one of the clustering algorithms that increase in popularity day by day. The intensive mathematical operations and the continuous increase of the data size while clustering on large data using the K-means algorithm prevent the algorithm from operating at high performance. Therefore, the K-means algorithm that works on large data needs to be implemented on very fast hardware. FPGAs capable of parallel processing can be mathematically processed much faster than traditional processors. Therefore, realization of algorithms that require intensive mathematical computations such as K- means using FPGAs is of great importance for the performance of applications.
In this study, an architecture is designed on the FPGA for the K-means algorithm and the accuracy and efficiency of the generated architecture are compared with the software applied in the standard processor and the performance is tested. When the results are examined, it is seen that the FPGA gives an average of 100X faster results than the standard processor.
