PRTNets cold-start recommendation using pairwise ranking and transfer networks

Majority of recommendations engines use matrix factorization techniques to represent users and items as decomposed latent matrices. These matrices represent the collaborative filtering domain, where recommendations is done by consideration of the users' rating of items. The user and latent matr...

Full description

Bibliographic Details

Main Author:	Valerio, Dylan (Author)
Other Authors:	Naval, Prospero C. Jr (adviser.)
Format:	Thesis
Language:	English
Subjects:	Recommender systems (Information filtering) Factorization (Mathematics)