Interactive user intent modeling for eliciting priors of a normal linear model

Iiris Sundin, Luana Micallef, Pekka Marttinen, Muhammad Ammad-Ud-Din, Tomi Peltola, Marta Soare, Giulio Jacucci, Samuel Kaski

Research output: Contribution to conferencePaperScientific


In this extended abstract, we present a novel interactive method to elicit the tacit
prior knowledge of an expert user for improving the accuracy of prediction models.
The main component of our method is an interactive user intent model that models
the domain expert’s knowledge on the relevance of different features for a prediction task. The user intent model selects the features on which to elicit user’s knowledge sequentially, based on the earlier user input. The results of a feasibility study show that our method improves prediction accuracy, when predicting the relative citation counts of scientific documents in a specific domain.
Original languageEnglish
Publication statusPublished - 2016
MoE publication typeNot Eligible
EventFuture of Interactive Learning Machines - Barcelona, Spain
Duration: 5 Dec 201610 Dec 2016


ConferenceFuture of Interactive Learning Machines
Abbreviated titleFILM


Dive into the research topics of 'Interactive user intent modeling for eliciting priors of a normal linear model'. Together they form a unique fingerprint.

Cite this