Zhao Song

Title: Innovations of Zhao Song in Secure Distributed Deep Learning

Introduction

Zhao Song is an accomplished inventor based in Princeton, NJ (US). He has made significant contributions to the field of deep learning, particularly in the area of secure and robust distributed systems. His innovative approach to encrypting data for neural networks showcases his expertise and forward-thinking mindset.

Latest Patents

Zhao Song holds a patent titled "System and method for secure and robust distributed deep learning." This patent describes a method for encrypting image data for a neural network, which includes mixing the image data with other datapoints to form mixed data. Additionally, it involves applying a pixel-wise random mask to the mixed data to create encrypted data. The patent also outlines a method for encrypting text data for natural language processing, which includes encoding each text datapoint via a pretrained text encoder, mixing the encoded datapoints, and applying a random mask to form encrypted data. This innovative approach enhances the security and robustness of deep learning systems.

Career Highlights

Zhao Song is affiliated with Princeton University, where he continues to advance research in deep learning and data security. His work has garnered attention for its practical applications and potential to improve the safety of neural networks.

Collaborations

Zhao has collaborated with notable colleagues such as Sanjeev Arora and Kai Yi Li, contributing to a rich environment of innovation and research at Princeton University.

Conclusion

Zhao Song's contributions to secure distributed deep learning represent a significant advancement in the field. His innovative methods for encrypting data not only enhance security but also pave the way for future developments in neural network applications.