超表面光学计算实现与应用研究

Research on Realization and Application of Metasurface Optical Computing

大数据时代的海量数据处理任务和人工智能技术的发展产生了对算力的爆炸式需求。传统电子计算系统由于受到集成电路技术和本身的架构的制约,难以在算力和计算速度上有质的提高。算力不足已经成为人工智能技术发展的瓶颈,发展新的计算范式从基础研究转变为一种紧迫的现实需求。光学计算系统高速、高并行和低功耗等特性契合了对理想计算范式的要求。本文通过基于超表面的光学计算系统设计,利用其自由度高、功能强大的特点,实现诸如卷积、微分、积分等数学运算,并在边缘检测、模式识别、图像处理等人工智能和深度学习领域开展应用比较研究,从而论证得到超表面的光学计算系统替代电子计算系统的可行方案。

The massive data processing tasks in the era of big data and the development of artificial intelligence technology have produced explosive demand for computility.Due to the constraints of integrated circuit technology and Von Neumann architecture,it is difficult for traditional electronic computing systems to improve their computility and computing speed qualitatively.The lack of computility has become the bottleneck of the development of artificial intelligence technology,and the focus of new computing paradigms has changed from basic research to an urgent actual requirement.The characteristics of optical computing systems such as high speed,high parallelism,and low power consumption meet the requirements of the ideal computing paradigm.In this paper,the optical computing system based on metasurface with high design freedom and powerful functions,which conveniently to realize mathematical operations including convolution,differentiation,and integration.We carry out comparative studies in the field of artificial intelligence and deep learning,such as edge detection,pattern recognition,image processing,etc.The feasibility of metasurface optical computing system instead of electronic computing system is demonstrated.