Yishi Xu

Title: Yishi Xu and Graph-Aware Incremental Learning for Recommender Systems.

Introduction

Yishi Xu is an inventor associated with Montreal, Canada, and Huawei Technologies Co., Ltd. His patent work addresses how recommender systems can adapt to new graph data while retaining knowledge learned from existing data.

Latest Patents

The provided information lists 1 patent for Xu: “Graph Structure Aware Incremental Learning for Recommender System.” The invention describes a system and method for training a recommender system that makes recommendations using a bipartite graph of user and item nodes.

The method initializes an updated graph neural network with parameters from an existing model and trains it on a new graph. A loss function distills knowledge from node embeddings generated by the existing model on the existing graph. The recommender system then replaces its existing model with the updated model. This approach is designed to incorporate new graph information while preserving useful knowledge from prior training.

Career Highlights

Xu’s listed patent connects his work at Huawei Technologies Co., Ltd. with machine learning and recommender-system development. Its focus is incremental learning for graph neural networks, a technique intended to help recommendation models adapt as user-item data changes.

Collaborations

The provided information identifies Huifeng Guo as Xu’s coworker. It does not specify whether she is a co-inventor or collaborator on the listed patent.

Conclusion

Yishi Xu’s listed invention explores a knowledge-distillation approach to updating graph-based recommender systems, reflecting an effort to help these systems learn from new data while retaining information from existing models.