Syllabus
Modeling customer preferences and querying with preferences
Introduction, motivation, challenges and use cases of customer preferences,
Lean startup model,
LMPM - Linear Monotone Preference Model, uniqueness of LMPM representation.
Top-k algorithms - querying/searching with preferences
Fagin’s monotone model of customer preferences and algorithms for computing top-k.
Theoretical optimality of threshold algorithm and practical experiments, a multi-customer model
Learning customer preferences
Problem of learning (acquisition) of customer preference
Learning customer preferences
Various metrics for evaluation the quality of models
Formal framework for transferability of preference models, connections to economical and optimization models
Mathematical Fuzzy Datalog - Preferential Datalog
Preferential logic as a language for modeling of preferences, many valued modus pones and its correctness
Procedural and declarative semantics of preferential Datalog without negation and with recursion, correctness
Fixpoint for preferential Datalog and computability of the minimal model
Theorem on approximate completeness of preferential Datalog
Annotation
We are interested in the process which governs customer’s interface action and system response of an e-shop.
We learn: to create and evaluate customer preference models based on some business models; to effectively find top-k answers; a domain calculi for these.
Labs are composed of reporting on current achievements, preference learning, a project of a virtual Lean Startup and customer imitation via a social network.