Acceleration of Deep Neural Network Training Using Field Programmable Gate Arrays

Figure 2

Max pool, taken form [19].