イベント名
IEICE ICETC 2021
発表年月日
2021/12/01
タイトル
Machine Learning for Intelligent Wireless Communications
講演者
Tomoaki Ohtsuki(Keio University), 
抄録
With the rapid development in AI and machine learning (ML), future wireless communication systems will have much more intelligence. For problems that can be accurately modeled, traditional algorithms show good performance and efficient solutions on partially convex problems. However, for some non-convex problems, existing algorithms usually obtain more efficient solutions while allowing a certain performance loss. At this time, ML is used to mine the parameter information of the known structure algorithm from the obtained data samples, to improve the convergence speed of the algorithm and the performance of the algorithm. However, a large number of 5G and B5G scenarios cannot be modeled by exact mathematical models such as some efficiency and latency optimization problems, resulting in the inability to obtain an accurate algorithm structure. In this case, the artificial neural networks (ANNs) including deep neural network (DNN), convolutional neural network (CNN), and so on are used to parameterize the model or algorithm. These methods that obtain model or algorithm features from massive amounts of data rather than based on pre-established rules are generally called data-driven. This talk introduces ML based intelligent wireless communications, mainly focusing on the research and application of ML in the physical layer.

Webinar視聴