큰 규모의 인공신경망 응용을 위한 인공신경망 하드웨어 아키텍쳐

Many artificial neural networks (ANN) have been implemented for various applications. However, handling large networks is still challenging because of its huge power and area overhead. In this paper, we propose the novel architecture which can support large-scale networks such as image processing as well as other various applications by dynamically decreasing the number of hardware neurons. While its learning capability is obtained from software using back propagation algorithm, the proposed architecture is exemplified by optical character recognition (OCR) system and implemented using 65nm CMOS process.