Duanshun Li

Title: Duanshun Li: Innovator in Neural Network Compression

Introduction

Duanshun Li is a prominent inventor based in Plainsboro, NJ (US). He has made significant contributions to the field of neural networks, particularly in the area of video encoding and decoding. His innovative approach has led to the development of a unique method that enhances the efficiency of neural network models.

Latest Patents

Duanshun Li holds a patent for "Methods and apparatus for kernel tensor and tree partition based neural network compression framework." This patent describes a method of encoding or decoding a video that involves a current picture, a first reference picture, and a weight tensor associated with a trained neural network (NN) model. The method includes generating kernel tensors, input channels, and output channels linked to the weight tensor. Each kernel tensor is associated with various types, including layer type and tree partition type. The process also involves creating tree partitions for coding tree units, coding units, prediction units, and transform units, ultimately leading to a compressed representation of the trained NN model.

Career Highlights

Duanshun Li is currently employed at Interdigital VC Holdings, Inc., where he continues to push the boundaries of technology in neural networks. His work has garnered attention for its innovative approach to video processing and neural network efficiency.

Collaborations

Throughout his career, Duanshun has collaborated with notable colleagues, including Hua Yang and Dong Tian. These partnerships have contributed to the advancement of his research and the successful development of his patented technologies.

Conclusion

Duanshun Li's contributions to neural network compression represent a significant advancement in the field of video encoding and decoding. His innovative methods and collaborative efforts continue to influence the technology landscape.