公开(公告)号：US12141674B2

公开(公告)日：2024-11-12

申请号：US17210034

申请日：2021-03-23

Applicant: Beijing Baidu Netcom Science And Technology Co Ltd

Inventor： Yixuan Shi ,  Mingyang Dai ,  Zixiang Liu

IPC: G06N3/04 ,  G06N3/08 ,  H04L67/50

Abstract: The present application discloses a method for generating a user interactive information processing model and a method for processing user interactive information, relates to the technical field of graph neural networks, and particularly relates to a user interactive information processing technology. The method comprises the following steps: determining node representations of network nodes of each layer in a graph neural network in an iterative manner, wherein the node representation of a single-layer node is obtained by representations of neighbor nodes of the single-layer node; adding the attribute feature of each node in each layer of network nodes to the node representation of each layer of network nodes to obtain a graph neural network to be trained; and training the graph neural network to be trained based on existing user interactive information, to obtain a user interactive information processing model, wherein the user interactive information processing model comprises adjusted node representation.

公开(公告)号：US20210209446A1

公开(公告)日：2021-07-08

申请号：US17210034

申请日：2021-03-23

Applicant: Beijing Baidu Netcom Science And Technology Co Ltd

Inventor： Yixuan Shi ,  Mingyang Dai ,  Zixiang Liu

IPC: G06N3/04 ,  G06N3/08 ,  H04L29/08

Abstract: The present application discloses a method for generating a user interactive information processing model and a method for processing user interactive information, relates to the technical field of graph neural networks, and particularly relates to a user interactive information processing technology. The method comprises the following steps: determining node representations of network nodes of each layer in a graph neural network in an iterative manner, wherein the node representation of a single-layer node is obtained by representations of neighbor nodes of the single-layer node; adding the attribute feature of each node in each layer of network nodes to the node representation of each layer of network nodes to obtain a graph neural network to be trained; and training the graph neural network to be trained based on existing user interactive information, to obtain a user interactive information processing model, wherein the user interactive information processing model comprises adjusted node representation.