Mingen Shih

Title: Mingen Shih: Innovator in Neural Network Acceleration

Introduction

Mingen Shih is a prominent inventor based in Hsinchu, Taiwan. He has made significant contributions to the field of deep learning, particularly in the optimization of neural network models. His innovative work focuses on enhancing the efficiency of neural networks through advanced compilation techniques.

Latest Patents

Mingen Shih holds a patent for "Compiling asymmetrically-quantized neural network models for deep learning acceleration." This patent describes a system that receives a neural network model containing asymmetric operations. Each of these operations includes fixed-point operands that are asymmetrically-quantized from corresponding floating-point operands. The system compiles the asymmetric operation into a symmetric operation, which includes a combined bias value. The compiler computes this combined bias value as a constant by merging input and output points of the asymmetric operation. Ultimately, the system generates a symmetric neural network model for inference hardware to execute in fixed-point arithmetic. This innovation is crucial for improving the performance of deep learning applications.

Career Highlights

Mingen Shih is currently employed at MediaTek Corporation, a leading global semiconductor company. His work at MediaTek focuses on advancing technologies that drive the future of connectivity and multimedia experiences. His expertise in neural networks positions him as a key player in the company's research and development efforts.

Collaborations

Mingen has collaborated with notable colleagues, including Chih-Wen Goo and Pei-Kuei Tsung. These collaborations have fostered a creative environment that encourages innovation and the development of cutting-edge technologies.

Conclusion

Mingen Shih's contributions to the field of deep learning and neural network optimization are noteworthy. His patent on compiling asymmetrically-quantized neural network models showcases his innovative approach to enhancing deep learning acceleration. Through his work at MediaTek Corporation and collaborations with talented colleagues, Mingen continues to push the boundaries of technology.