Wei Zuo

Title: Wei Zuo - Innovator in Neural Network Graph Partitioning

Introduction

Wei Zuo is an accomplished inventor based in Cupertino, CA (US). He has made significant contributions to the field of neural networks, particularly in optimizing their performance on resource-constrained hardware systems. His innovative approach has the potential to enhance the efficiency of various applications that rely on neural networks.

Latest Patents

Wei Zuo holds a patent for a "Neural networks graph partitioning system and method for the same." This invention discloses a graph partitioning system designed to run neural networks on hardware with limited resources. The system partitions a neural network graph into a series of sub-graphs, allowing multiple sub-graphs to be executed on available hardware subsystems. It utilizes a cost function based on estimated computation time and memory bandwidth of the partitioned sub-graphs. The system is a cycle estimation model of hardware that can operate quickly and parameterize memory latency. It supports heterogeneous partitioning for various types of accelerators, including CPU, GPU, and ASIC.

Career Highlights

Wei Zuo is currently associated with Black Sesame Technologies Inc., where he continues to innovate in the field of neural networks. His work focuses on improving the efficiency and performance of neural network applications, making them more accessible for various hardware configurations.

Collaborations

Wei has collaborated with notable colleagues, including Qiang Zhang and Chenhao Fang. Their combined expertise contributes to the advancement of technology in the field of neural networks.

Conclusion

Wei Zuo's contributions to neural network graph partitioning represent a significant advancement in the field of artificial intelligence. His innovative solutions are paving the way for more efficient use of hardware resources in neural network applications.