Flexible Modularized Artificial Neural Network Implementation on FPGA

This work presents a parameterized and modularized approach for the implementation of artificial neural network (ANN) on a field-programmable gate array (FPGA). The design investigates how to efficiently model an ANN that is easily adoptable to various applications with least modifications to the hardware description language (HDL). The Verilog HDL has been used to model the network. Fixed point precision and activation function implementations have been investigated to monitor FPGA resource utilization for different network topologies. The network implemented has been adopted to different applications including handwritten recognition of the MNIST dataset. This work shows that a well modularized network is easily adoptable for different applications hence helping take advantage of the re-configurability of FPGAs.