Syllabus
User preferences
- Introduction, motivation, challenges and use-cases of user preferences
- Modelling / expressing user preferences, types of user feedback, Linear Monotone Preference Model
- Learning user preferences, feedback interpretation, aggregating preferences & fuzzy logic
- Applications for user preferences, recommender systems, personalized search, challenge-response model
Advanced topics from recommender systems
- Fairness and proportionality in recommender systems
- Multicriterial optimization & evaluation in recommender systems
- Dynamic recommender systems: multi-armed bandits & reinforcement learning
- Unbiased evaluation, inverse propensity, feedback loops problem
- Deep learning for heterogeneous recommender systems
Annotation
This course focus on deeper understanding of user preferences/needs/requirements. The problem depends both on the user seniority, visits frequency or the particular domain in question.
We will focus e.g. on proportionality, preference drift, multicriteriality, unbiased evaluation and on algorithms capable to learn and recommend from such data. We will also focus on a wider context of preference interpretation, e.g. during search (strict/fuzzy preference, graphical interpretation).
Labs will mainly focus on refering about recent papers and a virtual "Lean startup" project.