Technical Program

Paper Detail

Paper: PS-2B.43
Session: Poster Session 2B
Location: Symphony/Overture
Session Time: Friday, September 7, 19:30 - 21:30
Presentation Time:Friday, September 7, 19:30 - 21:30
Presentation: Poster
Publication: 2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania
Paper Title: Improving associative learning studies with Bayesian experimental design
Manuscript:  Click here to view manuscript
Authors: Filip Melinscak, Dominik R. Bach, University of Zurich, Switzerland
Abstract: Associative learning theory has a rich tradition of formalized computational models. With increasingly complex and flexible models, it is becoming difficult to intuit experiments that can empirically distinguish between them. To address this issue, we propose a quantitative approach. Using the formalism of Bayesian experimental design, we tune experimental variables to maximize the utility of the experiment, i.e., to best discriminate computational models. We demonstrate the proposed method on two scenarios from the literature on models of classical conditioning. In both cases, optimized designs substantially outperform existing canonical designs in simulations: the odds of recovering the true model increase 15 times in one scenario, and 43 in the other. These results suggest that formally optimizing associative learning studies has potentially large benefits in terms of more accurate model selection.