Acceleration of Deep Neural Network Training Using Field Programmable Gate Arrays

Figure 6

Throughput with difference data quantization on FPGA.