Abstract:Collaborative filtering by the learning manifold alignments provides a new way for cross system personalization. It uses similarity between users to compute reconstruction weights. However, inaccurate similarity often leads to inaccurate weights and poor recommendation quality. By combining the topology and geometry structures of data set to calculate weight matrix, an improved collaborative filtering algorithm based on manifold alignments is proposed. The proposed algorithm removes the effect of similarity error on recommendation quality effectively. The experimental result indicates that the improved algorithm has better recommendation quality than that of the original algorithm.
[1] Mehta B, Hofmann T. Cross System Personalization and Collaborative Filtering by Learning Manifold Alignments // Proc of the 29th Annual German Conference on Artificial Intelligence. Bremen, Germany, 2006: 244-259 [2] Ham J, Lee D, Saul L. Semisupervised Alignment of Manifolds // Proc of the 10th International Workshop on Artificial Intelligence and Statistics. Barbados, West Indies, 2005: 120-127 [3] Tan Lu, Yi Dongyun, Feng Guozhu, et al. Local Invariable Projection. Natural Sciences Progress, 2004, 14(3): 282-287 (in Chinese) (谭 璐,易东云,冯国柱,等.局部不变投影.自然科学进展, 2004, 14(3), 282-287) [4] Belkin M, Niyogi P. Laplacian Eigenmaps for Dimensionality Reduction and Data Representation. Neural Computation, 2003, 15(6), 1373-1396 [5] Breese J S, Heckerman D, Kadie C M. Empirical Analysis of Predictive Algorithms for Collaborative Filtering // Proc of the 14th Conference on Uncertainty in Artificial Intelligence. San Francisco, USA, 1998: 43-52