摘要
在线食谱分享流行,食物推荐变得关键。基于现有推荐系统存在的问题:(1)忽略食材关联,影响个性化推荐;(2)未考虑用户对卡路里的偏好,本文提出自监督的卡路里感知异构图网络(SCHGN),建模食材关系并融入卡路里信息。该模型构建用户、食谱、食材和卡路里的异构图,并揭示它们之间的复杂联系。其次,通过自监督学习探索食材共现性,使用分层传递和注意力机制捕捉用户对卡路里的动态喜好。食物推荐基于用户的卡路里感知与食谱表示的相似性。实验证明此方法有效。
Online recipe sharing is popular,and food recommendation becomes crucial.Existing recommendation systems have problems:(1)ignoring the association between ingredients,affecting personalized recommendations;(2)not considering users'preferences for calories.We propose a self-supervised calorie-aware heterogeneous graph network(SCHGN),modeling the relationship between ingredients and incorporating calorie information.This model builds a heterogeneous graph of users,recipes,ingredients,and calories,revealing their complex connections,exploring ingredient co-occurrence through self-supervised learning,and using hierarchical propagation and attention mechanisms to capture users'dynamic preferences for calories.Food recommendations are based on the similarity between users'calorie-aware representations and recipe representations.Experiments prove the effectiveness of this method.
作者
曾子文
袁操
ZENG Ziwen;YUAN Cao(School of Mathematics and Computer Science,Wuhan Polytechnic University,Wuhan 430023,China)
出处
《武汉轻工大学学报》
CAS
2024年第3期82-89,105,共9页
Journal of Wuhan Polytechnic University
关键词
食物推荐
食谱热量
异构图
自监督学习
food recommendation
recipe calories
heterogeneous graph
self-supervised learning
