FPGA Trax Solver based on a neural network design

摘要

For research related to field programmable gate arrays (FPGAs), various FPGA game solvers have been developed recently. The FPGA game solver can achieve up to 1,000 times faster processing speeds than processor-based operations. This paper presents one development of such FPGA game solvers: an FPGA Trax Solver based on a neural network design. The speed of the FPGA Trax Solver operations implemented on an FPGA (Arria II GX; Altera Corp.) is over 60 times greater than operation of C++ based software on a personal computer (Vostro 460; Dell Inc.).