Preference Elicitation [Conjoint Analysis]
Conjoint Analysis Market research: assess consumer’s preferences on homogenous class of products Approach: describe products in terms of attributes and levels [conjoint structure]. Example: Cars = (Max. Speed) x (Gas Mileage) Max. Speed = { 100 mph, 120 mph, 150 mph} Gas Mileage = { 20 mpg, 17 mpg, 13 mpg, 10 mpg}
Pairwise Comparsions Which car are you more likely to buy? § 150 mph 100 § 10 mpg 13 § 120 mph 100 § 17 mpg 20 Tradeoff!
Marketing Approach § Given a set of products X § Elicit consumer’s preferences from pairwise comparisons [simulates real choice tasks] § Only small [constant] number of questions per respondent § For each respondent value function v: X →[0, 1]
Optimizing Visualization Systems Which (volume) rendering shows more detail?
Optimizing Visualization Systems Which (volume) rendering do you like better?
Optimizing Visualization Systems Which (volume) rendering shows more detail?
Optimizing Visualization Systems Which (volume) rendering shows more detail?
Netflix Challenge
Netflix Challenge http: //www. netflixprize. com/index
Netflix Challenge
Netflix Challenge § Challenge: From given ratings predict rating of unrated movies. § Training data set: >100 million ratings from >480 thousand customers on ~18 thousand movies. § Test data: 2. 8 million customer/movie pairs with the ratings withheld. § Compare to Netflix’ predictor ‘Cinematch’ § Quality measure: root mean square error
Surface Reconstruction
Surface reconstruction
Surface Reconstruction