Po-Yuan Chen

Title: Po-Yuan Chen and Deep Learning Network Optimization.

Introduction

Po-Yuan Chen is an inventor based in Taipei, Taiwan, whose patent work addresses the optimization of deep learning networks. His listed patent count is 1 patent, and he is associated with Wistron Corporation.

Latest Patents

Chen’s listed patent is “Optimizing Method and Computing Apparatus for Deep Learning Network and Computer-Readable Storage Medium.” The invention describes obtaining a value distribution from a pre-trained model and identifying breaking points that divide the distribution’s range into sections. It then applies different quantization parameters to values in different sections. The method is intended to reduce accuracy loss during quantization.

Career Highlights

Chen’s documented innovation focuses on applying section-based quantization to deep learning network parameters. The patent identifies a computing apparatus and a computer-readable storage medium as additional forms of the disclosed technology.

Collaborations

Chen is associated with Wistron Corporation. Jiun-In Guo is identified as a coworker.

Conclusion

Po-Yuan Chen’s listed patent presents a method for quantizing deep learning network values with different parameters across sections, with the goal of reducing accuracy loss.