Unlock instant, AI-driven research and patent intelligence for your innovation.

Method for learning geometric decoupling representation based on geometric non-entangled variational automatic encoder

An autoencoder, geometric technology, applied in the information field, can solve problems such as low expressiveness

Pending Publication Date: 2021-11-30
BEIJING UNIV OF POSTS & TELECOMM
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

At the same time, some items may repeat, leading to loop structures (e.g., u 1 →i 1 → u 2 →i 2 → u 1 )
Therefore, separating this hybrid structure into a Euclidean space with limited expressiveness may lead to lower expressiveness

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Method for learning geometric decoupling representation based on geometric non-entangled variational automatic encoder
  • Method for learning geometric decoupling representation based on geometric non-entangled variational automatic encoder
  • Method for learning geometric decoupling representation based on geometric non-entangled variational automatic encoder

Examples

Experimental program
Comparison scheme
Effect test

experiment example

[0076] Experimental approach: We conduct experiments on four real-world datasets. Specifically, we use the AliShop dataset and three MovieLens datasets of different scales, namely MovieLens-100k, MovieLens-1M, MovieLens-20M. The AliShop dataset contains user-item interactions associated with seven categories of items from Alibaba's e-commerce platform Taobao. The MovieLens dataset describes ratings and free-text tagging activity on the MovieLens website. We binarize the MovieLens dataset, maintain ratings of 4 or higher, and retain users who have watched at least 5 movies.

[0077] The invented method is compared with two state-of-the-art graph collaborative filtering methods and two separation-based recommendation methods: (1) NGCF is a graph-based CF model that incorporates high-order connectivity of user-item interactions, (2) LightGCN is a CF recommendation model based on graph convolutional network, (3) DGCF is a separate CF model that learns the representation of diffe...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a method for learning geometric decoupling representation based on a geometric non-entangled variational automatic encoder, which tries to learn geometric decoupling representation for the first time, proposes a geometric decoupling variational automatic encoder model (GDVAE), and projects separated representations in different geometric spaces into a shared potential space. Therefore, a general measure can be utilized to calculate the proximity, features under different geometric features can be learned, more effective feature representation can be obtained in combination with different geometries, and the effectiveness of the proposed GDVAE model is proved through experimental results.

Description

technical field [0001] The invention relates to the field of information technology, in particular to a method for learning geometric decoupling representations based on geometric non-entangled variational automatic encoders. Background technique [0002] The rapid development of information technology has promoted the explosive growth of information and intensified the challenge of information overload. The goal of recommender systems is to alleviate information overload by recommending a small set of items for users to meet their individual interests. Learning representations that reflect user preferences primarily based on user behavior has been a central topic in recommender systems research. In recommender systems, user behavior is driven by the complex interaction of many underlying preference factors behind the user's decision-making process. Several works study user / item decoupled feature representations to reveal and disentangle these intent factors. [0003] Tra...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06F30/27G06F111/04G06F111/08
CPCG06F30/27G06F2111/04G06F2111/08
Inventor 石川王啸张依丁
Owner BEIJING UNIV OF POSTS & TELECOMM