排序方式: 共有123条查询结果,搜索用时 15 毫秒
121.
提出具有解耦能力的多通道图注意力社交推荐模型,该模型主要包括深度聚类模块、多通道图注意力聚合模块和评分预测模块。其中,深度聚类模块用于对用户和项目进行分组,并利用聚类结果将用户社交图和用户项目图拆分成多个用户社交子图及用户项目子图,以学习用户兴趣分组及用户对不同类项目的兴趣;多通道图注意力聚合模块学习不同子图对预测结果的注意力;评分预测模块将学习到的用户表示向量和项目表示向量输入多层感知机进行评分预测。在多个真实数据集上的实验结果表明:提出的方法优于其他社交推荐算法。与最新的用于社交推荐的图神经网络方法相比,在Ciao和Epinions数据集上,均方根误差分别降低了2.26%和2.07%,平均绝对误差分别降低了2.58%和3.06%。 相似文献
122.
Tamir Libel 《战略研究杂志》2020,43(5):686-709
ABSTRACT The strategic culture approach has been suffering from a prolonged theoretical stalemate, despite a surge in case studies, which culminated in the Johnston-Gray debate and subsequent schism. The present paper outlines a new approach designed to overcome this deadlock, and consists of three arguments. First, the three previous generations of strategic culture studies have failed to explain how strategic culture influences behaviour. Second, aligning strategic culture theory-building with discursive institutionalism offers a way to overcome this fundamental fallacy. Third, a research programme for strategic culture should draw on computational social science to enable it to present and test middle-range theories. 相似文献
123.