Meta-learning method and device of cold start recommendation system, equipment and storage medium

Abstract

The invention discloses a meta-learning and device of a cold start recommendation system, equipment and a storage medium. The cold start problem is solved through a meta-learning method of a model level and a heterogeneous information network of a data level, a meta-learning model is prevented from entering local optimum to a certain extent by adopting a memory enhanced meta-optimization method, and a recommendation result of a cold start recommendation system is obtained according to a recommendation model. According to the method, on the data level, related semantic information on data is enriched by constructing meta-paths; and in the model level, a memory capable of storing specific semantics is adopted for guiding a model with semantic parameter initialization, and a meta-optimization method is adopted for optimizing the method so as to achieve rapid adaptation.

Description

technical field [0001] The present application relates to the field of recommendation systems, in particular to a meta-learning method, device, device and storage medium for a cold-start recommendation system. Background technique [0002] Since the rapid development of mobile applications, recommendation systems have played an increasingly important role in the industry. The core is to solve the problem of user information overload, but it also brings many challenges. Although the recommendation system can be divided into several categories, the traditional recommendation method based on matrix factorization or the technology based on deep learning has achieved success, but an unavoidable problem in most recommendation systems is the cold start problem. Due to the lack of user-item interaction, it is often difficult for recommender systems to make accurate recommendations for new users. The cold start problem can be further divided into user cold start and item cold start,...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a meta-learning method, device, equipment and storage medium for a cold-start recommendation system, aiming to solve the technical problem that the prior art cannot stably perform meta-learning

Images