Photonic Tensor Convolution

Release Time：2022-11-15Number of visits：3502

Speaker:    Wenjia Zhang，Shanghai Jiao Tong University

Time:         13:00-15:00 , Nov .15

Location:   SIST 1A 200

Host:          Cheng Wang

 

Abstract:

Convolutional neural network is one of the most successful algorithms in the fields of computer vision and object detection. With the explosive bandwidth growth of high-definition images and videos, intelligent computing processors require higher computational capability with less power consumption. Photonic technology has inherent capability of coherent combination and multidimensional manipulation, and will become an inevitable approach to realize tensor convolution operations. In this talk, we introduce a tensor convolution operation method based on space-time duality. By revealing the mathematical and physical relationship between tensor convolution and multi-dimensional optical transformation, we will discuss three key problems including dimension reduction operation, design philosophy of optical chip for tensor convolution operation and synergy of analog and digital components for fully-functional convolutional neural network. The development of proposed architecture will foster new technology of transmission as computing, and promotes intelligent computing processors from matrix manipulation to tensor convolution.

 

Bio:

Wenjia Zhang is an associate professor of state key laboratory of advanced optical communication systems and networks, Shanghai Jiao Tong University and co-PI at Peng Cheng laboratory. He received the B.S.E degree and Ph.D from Beijing University of Posts and Telecommunications. During this period, he visited the lightwave research laboratory at Columbia University and was with MIT and then Singapore-MIT alliance for research and technology as a postdoc researcher, followed by industry experience at Finisar as a senior optical engineer. His research interests include intelligent optical interconnect and computing. Dr. Zhang has published more than 70 journal and conference papers, with 10 invited talks in the international conferences, and organized the optical computing forum of online program initiated by CSOE as the chairman. His work on photonic convolution neural network received Tingye Li Best Paper Award at 2022 Asia Communications and Photonics Conference. He has been PI of youth, general and key project of national science foundation of China and national key R&D program.