公开(公告)号：US20240022620A1

公开(公告)日：2024-01-18

申请号：US18361504

申请日：2023-07-28

Applicant: Huawei Technologies Co., Ltd.

Inventor： Jian Li ,  Han Su

IPC: H04L65/75 ,  H04L65/61

CPC classification number: H04L65/762 ,  H04L65/61

Abstract: The present technology discloses communication in a communications network. A source node receives a media file from a client device in the communications network. Contextual information in the file is obtained that is descriptive of content in the medial file. When a match to the media file is found, the contextual information is accessed from storage. Otherwise, the contextual information is generated by the source node by extracting attributes from metadata in the media file. The media file is then transmitted over a first data path, and the contextual information is transmitted over a second data path, where the first data path has a higher latency than the second data path such that the contextual information arrives at the destination node prior to the media file.

公开(公告)号：US20230289563A1

公开(公告)日：2023-09-14

申请号：US18320007

申请日：2023-05-18

Applicant: Huawei Technologies Co., Ltd.

Inventor： Jian Li ,  Han Su

IPC: G06N3/045

CPC classification number: G06N3/045

Abstract: A method of training a large neural network using a number of pre-trained smaller neural networks. Multiple pre-existing, pre-trained neural networks are used to create the large neural network using multi-level superposition. The pre-trained neural networks, each having a first number of multi-dimensional nodes, are each up-scaled to provide larger, sparse neural networks. The values of the larger, sparse neural networks are superpositioned into the larger neural network. The pre-trained neural networks may be created from publicly available, pre-trained neural networks. The larger neural network can be adapted for use in a different task by replacing and/or re-training one of the sub-networks used to create the large neural network.