Design and Simulation Fuzzy Neuro Generalized Learning Vector Quantization-PI (FNGLVQ-PI) on Field Programmable Gate Array (FPGA)
Ricky A. Daniel, Muhammad Anwar Ma’Sum, Grafika Jati, Wisnu Jatmiko · Advances in physics research/Advances in Physics Research · 2014
Classification is a machine learning technique that used widely for various applications.One of the classification algorithm is FNGLVQ that can be used to solve several types of classification problem in previous research.On the other side, Field Programmable Gate Array (FPGA) is an instrument currently used in many application especially for portable smart device.In this paper we will discuss thoroughly about design and simulation of FNGLVQ algortihm on FPGA.FNGLVQ is an artificial neural network based algorithm that can be used for several applications in previous researches.The design consists of two major phases, training and testing phase.The design was implemented in Xilinx ISE Project Navigator which is an integrated development environtment (IDE) to build the design into the actual Xilinx FPGA.The IDE also provides simulation feature for the design.In this research, we use Iris dataset taken from UCI Machine Learning database.Simulation result shows that this design reached 90.00%, 93.33%, 93.33%, 83.33%, 80.00%, 83.00%, and 86.67% accuracy respectively for epoch value 1, 2, 4, 8, 16, 32 and 64.As comparison, FNGLVQ implemented in MATLAB constantly reached 93.33% accuracy for those variation of epoch. .However, running time on the FPGA side is approximately twenty time faster than on MATLAB side