Acceleration of Deep Neural Network Training Using Field Programmable Gate Arrays

Figure 5

Throughout with difference data quantization with Caffe [CPU].