Da Zheng

Title: Da Zheng: Inventor of Distributed Machine Learning Technology

Introduction

Da Zheng is an inventor based in Fremont, California, whose patent work addresses the challenges of training machine learning models on large graph datasets. His listed patent is associated with Amazon Technologies, Inc.

Latest Patents

Zheng is named on one patent, 1 patent: “Distributed asynchronous training of machine learning models for large graphs using multiple resource types.” The invention describes obtaining a graph dataset at a machine learning service and training a model through multiple mini-batch iterations. Each iteration uses asynchronous tasks, including graph sampling on a CPU and model-parameter learning performed at least partly on a training accelerator. Tasks from different iterations can run in parallel, and the trained model is then stored.

Career Highlights

The patent reflects work on coordinating different computing resources to train models on large graphs. It combines CPU-based graph sampling with accelerator-assisted learning and asynchronous, parallel execution.

Collaborations

Zheng’s listed coworkers include Xiang Song and Chengru Yang. Their names are provided as collaborators in the supplied information.

Conclusion

Da Zheng’s patent focuses on making large-graph machine learning training more efficient through asynchronous tasks and coordinated use of CPUs and training accelerators.