A popular model of preference in the context of recommendation systems is the so-called ideal point model. In this model, a user is represented as a vector u together with a collection of items x N in a common low-dimensional space. The vector u represents the user's ideal point, or the ideal combination of features that represents a hypothesized most preferred item. The underlying assumption in this model is that a smaller distance between u and an item x j. In the vast majority of the existing work on learning ideal point models, the underlying distance has been assumed to be Euclidean. However, this eliminates any possibility of interactions between features and a user's underlying preferences.

Weaknesses: Update: The authors provided thoughtful feedback on noise considerations that addresses the main weakness mentioned below from my original review. Provided that this discussion on the noise model is incorporated into the final text, I think the paper will be substantially stronger. Additionally, the authors provide detailed feedback for other weaknesses that other reviewers brought up. Altogether, the additional discussion makes the paper much more thorough. In light of this, I am increasing my score of the paper from 7 to 8. One unresolved issue though: one of the reviewers pointed out that the ideal GPA of 4.06 is greater than 4 despite the GPA being normalized to a 4.0 scale.

All of the reviewers are in agreement that this is a very nice paper which deals with a hot problem in an interesting fashion, has very interesting evaluations, and should be accepted to the conference.

A popular model of preference in the context of recommendation systems is the so-called ideal point model. In this model, a user is represented as a vector u together with a collection of items x1 ... xN in a common low-dimensional space. The vector u represents the user's "ideal point," or the ideal combination of features that represents a hypothesized most preferred item. The underlying assumption in this model is that a smaller distance between u and an item xj indicates a stronger preference for xj. In the vast majority of the existing work on learning ideal point models, the underlying distance has been assumed to be Euclidean. However, this eliminates any possibility of interactions between features and a user's underlying preferences.