J. Moore et Nm. Botros, DESIGN AND IMPLEMENTATION OF HOPFIELD NEURAL-NETWORK USING VHDL STRUCTURAL MODELING, International journal of robotics & automation, 12(1), 1997, pp. 33-37
The authors present a technique for hardware realization of a discrete
Hopfield artificial neural network. The technique is fast and expanda
ble, and produces a VHDL code describing the specifics of the network'
s architecture. The user specifies the size of the network and the tra
ining data st the time of compilation. The design of the network is pu
rely structural VHDL that allows tight control over all primitives. A
synthesis program (Viewsynthesis) is used on the structural design to
realize a six-neuron Hopfield network on a Xc4000 Field Programmable A
rrays (FPGAs) chip. The network is trained to recognize simple seven-s
egment patterns. The authors have successfully tested the network.