论文部分内容阅读
针对跨空间数据相似度学习问题提出的跨空间相似度学习(CSAL)算法表现出了良好的性能,并已成功地应用于各类推荐系统中.但构建一个完善的推荐系统,其待处理的数据量常呈现大样本特征,而CSAL算法并不具备大样本快速处理能力.针对此不足,提出了跨空间相似度学习-最小包含球(CSAL-MEB)方法和跨空间相似度学习-核向量机(CSAL-CVM)快速方法.CSAL-CVM方法既具有渐近线性时间复杂度和空间复杂度的优点,同时又继承了CSAL的良好性能.相关实验亦验证了所提出方法的有效性.
The Cross-space Similarity Learning (CSAL) algorithm proposed for cross-space data similarity learning has shown good performance and has been successfully applied to various kinds of recommendation systems. However, a perfect recommendation system is constructed, which is to be processed However, the CSAL algorithm does not have the ability to process large samples quickly.Aiming at this problem, we propose a cross-space similarity learning-least inclusion sphere (CSAL-MEB) method and cross-space similarity learning-kernel vectors (CSAL-CVM) fast method.CSAL-CVM method not only has the advantages of asymptotic linear time complexity and space complexity, but also inherits the good performance of CSAL.Experimental results also verify the effectiveness of the proposed method.