Recommendation method and system based on de-smoothed graph convolutional neural network
The invention provides a recommendation method and system based on a smoothing graph removal convolutional neural network. The recommendation method comprises the following steps: obtaining an initial embedding vector of a user node and an initial embedding vector of an article node through a user-a...
Saved in:
Main Authors | , , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
21.10.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention provides a recommendation method and system based on a smoothing graph removal convolutional neural network. The recommendation method comprises the following steps: obtaining an initial embedding vector of a user node and an initial embedding vector of an article node through a user-article interaction graph as training samples; carrying out smooth convolution combination operation on the training sample through a multilayer graph convolutional neural network model based on a de-smoothing module to obtain a user embedding vector and an article embedding vector of each order; respectively carrying out layer combination on the obtained user embedding vectors and article embedding vector representations of all orders, and carrying out inner product operation on the final user embedding vectors and article embedding vectors obtained after layer combination to obtain prediction scores of the training samples; and training the multilayer graph convolutional neural network model based on the de-smooth |
---|---|
Bibliography: | Application Number: CN202210883824 |