"Paper title" Local latent Space Models for Top-n recommendation (KDD-2018)
"thesis Author"-evangelia Christakopoulou (University of Minnesota), George Karypis (University of Minnesota)
"Paper Link" Paper (9-pages//Double column)
Summary
The user's behavior is determined by their purchase, Viewing the preferences of the various aspects of a potentially interesting commodity is driven by the potential spatial method to model these aspects in the form of . Although these methods have been proven to deliver good results, the important aspects of different users (which he prefers) may be different. In many areas, there may be to capture this explicitly, we present several models that contain All users are concerned about hidden features implied features
In particular, we present two potential (implicit) spatial models:RGLSVD and SGLSVD, which combine an implicit set of features such as a global and a specific subset of users.
RGLSVD Model according to User our comments points pattern assigns the user to a different subset and then estimates a global and specific Subset of users of the Local models, the potential of these models (implied) Number of dimensions (number of implied features) may be different.
SGLSVD Model by adding implied The number of dimensions remains in these models, resulting in a global and specific Subset of users Local The model is estimated, but the user's grouping is optimized to achieve the best approximation.
Our experiments on different real-world datasets show that the proposed approach is significantly better than the most advanced potential (implicit) space-N recommendation method.
"Some reference URLs"
1, Http://www.kdd.org/kdd2018/accepted-papers/view/local-latent-space-models-for-top-n-recommendation
2, https://www-users.cs.umn.edu/~chri2951/publications.html
"RS" using local implicit space model for TOP-N recommendations