Hanxi Heinrich Jiang

Title: Hanxi Heinrich Jiang and His Work on Machine-Learning Label Bias.

Introduction

Hanxi Heinrich Jiang is an inventor based in Mountain View, California. His patent work addresses a challenge in machine learning: correcting bias in training labels while preserving the labels themselves.

Latest Patents

Jiang is associated with 1 patent, titled “Identifying and Correcting Label Bias in Machine Learning.” The invention describes systems and methods that re-weight training examples to reduce label bias. It considers how an agent may intend to label examples accurately while still introducing biases toward certain groups. Although a biased dataset contains only the resulting labels, the disclosed approach aims to correct the bias by adjusting the examples’ weights without changing those labels.

Career Highlights

Jiang works at Google Inc. His patent reflects work on methods for improving the reliability of machine-learning training data.

Collaborations

Jiang’s listed coworker is Ofir Nachum. The provided information does not specify their particular projects or the nature of their collaboration.

Conclusion

Hanxi Heinrich Jiang’s patent focuses on reducing label bias through intelligent re-weighting, offering an approach to improving machine-learning datasets without altering their recorded labels.