2019 Dissertation Prize
In various scientific fields, statistical models of interest are analytically intractable and inference is usually performed using a simulation-based method. However elegant these methods are, they are often painstakingly slow and convergence is difficult to assess. As a result, statistical inference is greatly hampered by computational constraints. However, for a given statistical model, different users, even with different data, are likely to perform similar computations. Computations done by one user are potentially useful for other users with different data sets. In this presentation, I recommend a pooling of resources across researchers to capitalize on this. More specifically, I propose to preemptively chart out the entire space of possible model outcomes in a prepaid database. Using advanced interpolation techniques, any individual estimation problem can now be solved on the spot. I will discuss prepaid databases created for several challenging models and demonstrate how they can be distributed through an online parameter estimation service. This method outperforms state-of-the-art estimation techniques in both speed (up to a 100,000-fold speed up) and accuracy, and is able to handle previously quasi inestimable models. Specifically I will demonstrate how this online parameter estimation technique is able to quasi instantaneously estimate the parameters of choice reaction time models without a tractable likelihood.
ABOUT THE SPEAKER
During his master mathematical engineering at the University of Leuven, Merijn Mestdagh specialized in machine learning, statistics and computer sciences. Subsequently, he prepared a PhD at the University of Leuven under the supervision of Francis Tuerlinckx and Stijn Verdonck. During his PhD, he applied the insights of machine learning to substantive research of emotions, aiming to facilitate and improve the mathematical study of affect, which resulted in publications in Nature Human Behavior, Scientific Reports and Psychological Methods.