Digital system for spiking neural network emulation
Eduard-Guillem Merino Mallorquí · RECERCAT (Consorci de Serveis Universitaris de Catalunya) · 2017
The present project is about the design, simulation and an experimentational test of a digital system in a single chip able to emulate the behavior of spiking neural networks, which is possible thanks to the use of mathematical models that emulate the behavior of these networks in the brain. A modular system has been proposed in order to provide the necessary flexibility and scalability for the simulation of different neural networks. At the same time the most flexible, simple and efficient option has been chosen in order to have a good performance without losing or reducing the necessary accuracy and exactitude for the emulation of the neural networks. The solution has been implemented by making use of different combinational blocks and totally synchronous flip-flops from a 100 MHz clock signal, besides, the description of the system was performed by using the high-level hardware description language VHDL. Finally, a neural network for pattern recognition has been implemented on a programmable logical device FPGA in order to demonstrate the correct operation of the digital system.