Incremental Methods in Collaborative Filtering for Ordinal Data
In modern collaborative filtering applications initial data are typically very large (holding millions of users and items) and come in real time.In this case only incremental algorithms are practically efficient. In this paper a new algorithm based on the symbiosis of Incremental Singular Value Decomposition (ISVD) and Generalized Hebbian Algorithm (GHA) is proposed. The algorithm does not require to store the initial data matrix and effectively updates user/item profiles when a new user or a new item appears or a matrix cell is modified. The results of experiments show how root mean square error (RMSE) depends on the number of algorithm’s iterations and data amount.
KeywordsCollaborative filtering singular value decomposition Generalized Hebbian algorithm sparse matrix large data ordinal data incremental data
- 2.Takacs, G., Pilaszy, I., Nemeth, B., Tikk, D.: Scalable Collaborative Filtering Approaches for Large Recommender Systems. The Journal of Machine Learning Research 10, 623–656 (2009)Google Scholar
- 3.Gorrell, G.: Generalized Hebbian Algorithm for Incremental Singular Value Decomposition in Natural Language Processing. In: Proceedings of EACL (2006)Google Scholar
- 4.Golub, G.: Matrix Computations, p. 728. Johns Hopkins University Press (1996)Google Scholar
- 5.Brand, M.: Fast online SVD revisions for lightweight recommender systems. In: SIAM International Conference on Data Mining (SDM) (2003)Google Scholar