Oddly enough FDR1 and Churchill were eighth cousins once removed2, if the researches of one genealogist3 are correct. Both men, it appears, can trace a common descent4 from a personage known as John Cooke who came to ...Oddly enough FDR1 and Churchill were eighth cousins once removed2, if the researches of one genealogist3 are correct. Both men, it appears, can trace a common descent4 from a personage known as John Cooke who came to America on the Mayflower. John married Sarah Warren; one of their daughters was the great-great-great-great-great-great-grandmother of Sara Delano, FDR’s mother, and another was a direct ancestress5 of Churchill’s American-born mother, Jennie Jerome.展开更多
现有的基于异质信息网络的推荐方法通过最大池化操作提取元路径语义信息时,未考虑元路径语义的整体性特征及特征冗余问题的影响。针对这些问题,提出了一种融合元路径与改进协同注意力的推荐模型MICA(Research on Recommendation Fusing ...现有的基于异质信息网络的推荐方法通过最大池化操作提取元路径语义信息时,未考虑元路径语义的整体性特征及特征冗余问题的影响。针对这些问题,提出了一种融合元路径与改进协同注意力的推荐模型MICA(Research on Recommendation Fusing Meta-Path and Improved Collaborative Attention)。该模型通过基于k-最大池化的协同注意力机制更深层次地学习用户和项目的协同注意力嵌入表达,缓解了最大池化操作在协同注意力机制中带来的特征丢失问题;基于提取的路径序列显著性特征和完整性特征通过注意力机制学习元路径的重要特征,得到元路径上下文嵌入表达,保留有效的元路径语义信息;利用注意力机制学习融合后的用户、项目协同注意力嵌入表达和元路径上下文嵌入表达,以减少冗余信息,实现top-N推荐。在两个真实数据集上的实验结果表明,MICA在三个评价指标上均优于其他模型,能够更好地提高推荐性能,有效地提取特征信息,缓解元路径特征提取不充分等问题。展开更多
文摘Oddly enough FDR1 and Churchill were eighth cousins once removed2, if the researches of one genealogist3 are correct. Both men, it appears, can trace a common descent4 from a personage known as John Cooke who came to America on the Mayflower. John married Sarah Warren; one of their daughters was the great-great-great-great-great-great-grandmother of Sara Delano, FDR’s mother, and another was a direct ancestress5 of Churchill’s American-born mother, Jennie Jerome.
文摘现有的基于异质信息网络的推荐方法通过最大池化操作提取元路径语义信息时,未考虑元路径语义的整体性特征及特征冗余问题的影响。针对这些问题,提出了一种融合元路径与改进协同注意力的推荐模型MICA(Research on Recommendation Fusing Meta-Path and Improved Collaborative Attention)。该模型通过基于k-最大池化的协同注意力机制更深层次地学习用户和项目的协同注意力嵌入表达,缓解了最大池化操作在协同注意力机制中带来的特征丢失问题;基于提取的路径序列显著性特征和完整性特征通过注意力机制学习元路径的重要特征,得到元路径上下文嵌入表达,保留有效的元路径语义信息;利用注意力机制学习融合后的用户、项目协同注意力嵌入表达和元路径上下文嵌入表达,以减少冗余信息,实现top-N推荐。在两个真实数据集上的实验结果表明,MICA在三个评价指标上均优于其他模型,能够更好地提高推荐性能,有效地提取特征信息,缓解元路径特征提取不充分等问题。