Jianyu Wang

Title: Innovations of Jianyu Wang in Adversarial Machine Learning

Introduction

Jianyu Wang is a prominent inventor based in Santa Clara, CA, known for his contributions to the field of machine learning, particularly in developing robust models against adversarial attacks. With a total of 4 patents, his work focuses on enhancing the security and reliability of machine learning systems.

Latest Patents

Wang's latest patents include innovative systems and methods for fast training of more robust models against adversarial attacks. One of his notable inventions is the Bilateral Adversarial Training (BAT), which involves perturbing both the image and the label to generate adversarial labels through closed-form heuristic solutions. This method effectively addresses issues such as label leaking and gradient masking, significantly improving state-of-the-art results in various experiments.

Another significant patent by Wang is the feature-scattering-based adversarial training approach. This method enhances model robustness by generating adversarial images through unsupervised feature scattering in the latent space, thus avoiding label leaking. Extensive experiments have demonstrated the effectiveness of this approach compared to conventional adversarial training methods.

Career Highlights

Jianyu Wang is currently employed at Baidu USA LLC, where he continues to push the boundaries of machine learning research. His work has garnered attention for its innovative approaches to adversarial training, making significant strides in the field.

Collaborations

Wang collaborates with Haichao Zhang, contributing to the advancement of their shared research interests in machine learning and adversarial robustness.

Conclusion

Jianyu Wang's innovative work in adversarial machine learning showcases his commitment to enhancing the security of machine learning models. His patents reflect a deep understanding of the challenges in the field and offer promising solutions for future advancements.