Fei Yang

Title: Fei Yang: Innovator in Deep Learning and Language Model Compression

Introduction

Fei Yang is a prominent inventor based in Hangzhou, China. He has made significant contributions to the fields of deep learning and language model compression. With a total of 3 patents, his work focuses on enhancing the efficiency and adaptability of machine learning frameworks.

Latest Patents

Fei Yang's latest patents include a method for adapting a deep learning framework to a hardware device based on a unified backend engine. This innovative method involves several steps, including adding the unified backend engine to both the deep learning framework and the hardware device, converting a computational graph into an intermediate representation, and compiling this representation to generate an executable object. The final steps involve running the executable object on the hardware device and managing the memory of the unified backend engine.

Another notable patent is a compression method and platform for pre-training language models based on knowledge distillation. This method designs a universal knowledge distillation strategy that focuses on feature migration from a teacher model to a student model. It emphasizes the ability of small samples to express features in the intermediate layers of the teacher model, guiding the student model through these features. The method also constructs a knowledge distillation approach based on self-attention cross and incorporates a linear transfer strategy based on Bernoulli probability distribution to facilitate knowledge transfer.

Career Highlights

Fei Yang has worked with notable companies such as Zhejiang Lab and Future Wei Technologies, Inc. His experience in these organizations has allowed him to develop and refine his innovative ideas in the realm of artificial intelligence and machine learning.

Collaborations

Fei Yang has collaborated with esteemed colleagues, including Hongsheng Wang and Haijun Shan. These partnerships have contributed to the advancement of his research and the successful implementation of his patented technologies.

Conclusion

Fei Yang's contributions to deep learning and language model compression demonstrate his innovative spirit and commitment to advancing technology. His patents reflect a deep understanding of machine learning frameworks and their applications in real-world scenarios.