Company Filing History:
1 out of 928 
1 out of 116 
Years Active: 2025
Title: Innovations of Teng Xi in Neural Architecture Search
Introduction
Teng Xi is an accomplished inventor based in Beijing, China. He has made significant contributions to the field of neural architecture search (NAS), particularly through his innovative methodologies that address existing challenges in the domain. His work has garnered attention for its potential to enhance the efficiency of network architecture searches.
Latest Patents
Teng Xi holds a patent titled "Neural architecture search via similarity-based operator ranking." This patent focuses on improving the supernet-based differentiable approach, which is popular for effectively sharing weights during the search process. However, it also addresses the issues of mismatch between architecture and weights caused by weight sharing, as well as the neglect of coupling effects among different operators. His methodology utilizes a similarity-based operator ranking based on statistical random comparison to approximate each layer's output in the supernet. This innovative approach prunes operators that cause minimal changes to feature distribution discrepancies and employs a fair sampling process to mitigate the Matthew effect observed in previous supernet approaches. He has 1 patent to his name.
Career Highlights
Teng Xi has worked with notable companies such as Baidu USA LLC and Baidu Times Technology (Beijing) Co. Ltd. His experience in these organizations has allowed him to refine his skills and contribute to cutting-edge research in artificial intelligence and machine learning.
Collaborations
Teng Xi has collaborated with esteemed colleagues, including Baopu Li and Yanwen Fan. Their joint efforts have further advanced the field of neural architecture search and have led to innovative solutions that address complex challenges.
Conclusion
Teng Xi's contributions to neural architecture search demonstrate his commitment to innovation and excellence in the field. His methodologies not only enhance the efficiency of network architecture searches but also pave the way for future advancements in artificial intelligence.