On May 12, 2021, Associate Professor Xue Feng was invited to give a presentation entitled "All-Optical Neural Network with Programmable Linear Transformation" in “Conference on Lasers and Electro-Optics (CLEO) 2021, Applications & Technology Topical Reviews”, Session of “Integrated Photonics in Neural Networks”. In our previous works, programmable arbitrary linear optical operations have been demonstrated on discrete phase-coherent spatial modes with dimensionality of 24. Thus, in this work, we proposed and demonstrated a programmable optical neural network (ONN) scheme for various image identification tasks. The dimensionality of programmable optical matrix operations is 30´10/ 37´10 for the task of MNIST- digit/fashion, while the accuracy is 85%/81%, respectively.

CLEO is a world-known conference in the field of optoelectronics. Due to COVID-19, the CLEO2021 was held in an all-virtual, web conference format including more than 2000 conference Sessions. Over 200 invited speakers covered 29 different topic categories have shared their innovative idea and opinion. A large number of researchers and relevant practitioners all over the world have been attracted to attend this conference.