NEC develops learning-based analysis technology for local 5G: Page 2 of 2

May 13, 2020 //By Jean-Pierre Joosting
NEC develops learning-based analysis technology for local 5G
NEC Corporation has developed a learning-based communications analysis technology for private radio communications networks, including local/private 5G networks (local 5G). This technology ensures that the quality of communications is constantly sufficient, even without the assistance of a network specialist, avoiding quality degradation caused by congestion and competition.

NEC will develop this technology not only for local 5G, but also as a technology that facilitates the optimization of networks operated by a wide range of businesses, thereby contributing to the simplicity and quick realization of flexible and secure communications infrastructure.

A key feature of the new technology is the use of hierarchical communications analysis to enable  high-precision in real-time analysis. Conventionally, variations in communications traffic caused by a  mixture of application types, such as video, still image, and text data, and the effects of wireless communications, take time to analyze making the analysis less accurate. This new technology uses hierarchical clustering to first identify the wireless communications among the current communications traffic and to then identify the application types. This delivers highly accurate real-time analysis of communications systems.

Further, the technology learns autonomously and tracks changes in usage conditions automatically. Previously, the method for learning required enormous amounts of training data every time there was a change in the installation environment and the use-case of the communications system. This new method used by NEC is based on unsupervised learning. The technology automatically updates model parameters based on the similarity between the past and the most recent model, allowing changes in usage conditions to be tracked.

Part of this technology results from joint research between NEC and Professor Akihiro Nakao of The University of Tokyo, Interfaculty Initiative in Information Studies.

Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.