This inventor holds 1 USPTO granted patent and 3 published patent applications. Top assignee: Microsoft Technology Licensing, LLC. Active years: 2026.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile
Company Filing History:
Years Active: 2026
Title: Yabin Liu: Inventor in Neural Network Adaptation
Introduction
Yabin Liu is an inventor based in Kirkland, Washington. His patent work concerns adapting neural network models for specialized domains while preserving the original model weights.
Latest Patents
Liu is associated with 1 patent, titled “Low-Rank Adaptation of Neural Network Models.” The invention describes adding trainable low-rank factorization matrices to the weight matrices of multiple neural network layers to create a domain-specific model. The model is then trained on domain-specific data without modifying the base model’s weight matrices.
Career Highlights
Liu is associated with Microsoft Technology Licensing, LLC. His patent reflects an approach to adapting neural networks that keeps the base model weights unchanged during domain-specific training.
Collaborations
Liu’s listed coworkers include Weizhu Chen and Jingfeng Hu.
Conclusion
Yabin Liu’s patent focuses on efficient domain adaptation of neural network models through trainable low-rank matrices, and is associated with his work at Microsoft Technology Licensing, LLC.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile