We report top-N recommendation results on benchmark dat- asets using a simplified variant of Sparse Linear Methods (SLIM) which reduces the problem to be solved to a sim- ple linear regression: this allows the use of stochastic gra- dient descent (SGD) in place of more expensive optimiza- tion techniques. Our results suggest that recommendations generated in this way are as good or better than those pro- duced by the more computationally expensive method, as well as clearly superior to those from recent model-based preference prediction methods. We also introduce a simple method for tuning the regularization constants used during learning, which together with the use of SGD makes this approach highly practical for deployment in real world rec- ommender systems
CITATION STYLE
Levy, M., & Jack, K. (2013). Efficient Top-N Recommendation by Linear Regression. In Large Scale Recommender Systems Workshop in RecSys’13.
Mendeley helps you to discover research relevant for your work.