Acceleration of Deep Neural Network Training Using Field Programmable Gate Arrays

Figure 7

Energy efficiency with difference data quantization with Caffe [CPU].