WIRELESS NETWORK ENERGY SAVING WITH GRAPH NEURAL NETWORKS

    公开(公告)号:US20240023028A1

    公开(公告)日:2024-01-18

    申请号:US18470901

    申请日:2023-09-20

    CPC classification number: H04W52/223 G06N3/08

    Abstract: The present disclosure discusses network energy savings (NES) machine learning (ML) models that predict NES parameters used to adjust control parameters of respective network nodes in a wireless network, wherein the NES parameters can be used by the respective network nodes to adjust their control parameters, such that the wireless network realizes or achieves NES as a whole. The wireless network is represented as a graph with heterogeneous vertices that represent corresponding network nodes and edges that represent connections between the network nodes. The NES ML model comprises a graph neural network (GNN) and a fully connected neural network (FCNN). The GNN may be a graph convolutional neural network or a graph attention network. The FCNN may be a multi-layer perceptron, a deep neural network, and/or some other type of neural network. Other embodiments may be described and/or claimed.

    Two-stage spatial compression method

    公开(公告)号:US10211906B1

    公开(公告)日:2019-02-19

    申请号:US15928188

    申请日:2018-03-22

    Abstract: A wireless communication device comprises at least one antenna array configured to receive a radio signal; a first compressor configured to determine a first compression operator based on a predetermined set of receive configurations for the antenna array, and to apply the first compression operator to the radio signal to generate a first compressed signal; a second compressor configured to apply a second compression operator to the first compressed signal to generate a second compressed signal; a channel estimator configured to determine a channel covariance of the radio signal, determine at least one eigenvector of the channel covariance to obtain a second compression operator, and provide the second compression operator to the second compressor; and a baseband processor configured to receive the second compressed signal from the second compressor and to process the second compressed signal to obtain output data.

Patent Agency Ranking