We apply our approach to solve the calibration problem usingthe MovieLens dataset and compare our results with Steck’s state-of-the-art heuristic algorithm.
This paper contributes to the literature bysuggesting an optimization model that can calibrate the recom-mendation lists and maintain a good level of accuracy at the sametime.
we propose using adierent distance measure and show it has experimental and in-tuitive advantages.
Our optimization model is scalable. We solve objective func-tion (10) for every user separately, so the number of users do nothinder our model. We can solve our model with very large item setsby only considering items that have the highest estimated utility(e.g., top 1000 items) which in practice works eciently
Our goal is to compare Calib-Opt with Steck’s algorithm
we use the Gurobi software
For the calibration problem, Steck uses the KL-distance of distribu-tions
Calib-Opt and Steck’s heuristic
Glasp is a social web highlighter that people can highlight and organize quotes and thoughts from the web, and access other like-minded people’s learning.