Présentations à venir
Mercredi 27 octobre 2021 à 15h - Amphithéâtre - LaBRI Bât A30
Théo Matricon - M2F - LaBRI - Université de Bordeaux
Titre : Statistical Comparison of Algorithm PerformanceThrough Instance Selection
Résumé : Empirical performance evaluations, in competitions and scientific publications, play a major role in improving the state of the art in solving many automated reasoning problems, including boolean satisfiability (SAT), constraint programming and Bayesian network structure learning (BNSL). To empirically demonstrate the merit of a new solver usually requires extensive experiments, with computational costs of CPU years. This not only makes it difficult for researchers with limited access to computational resources to test their ideas and publish their work, but also consumes large amounts of energy. We propose an approach for comparing the performance of two algorithms: by performing runs on carefully chosen instances, we obtain a probabilistic statement on which algorithm performs best, trading off between the computational cost of running algorithms and the confidence in the result. We describe a set of methods for this purpose and evaluate their efficacy on diverse datasets from SAT, CSP and BNSL. On all these datasets, most of our approaches were able to choose the correct algorithm with about 95% accuracy, while using less than a third of the CPU time required for a full comparison; the best methods reach this level of accuracy within less than 15% of the CPU time for a full comparison.
Historique des présentations
Mercredi 13 octobre 2021 à 15h - Salle 178 - LaBRI Bât A30
Vincent Martin - Image et Son - LaBRI - Université de Bordeaux
Titre : How (and how not) to design a corpus for machine learning in psychiatry ?
Résumé : While the global effort in machine learning is focused on building ever more complex learning models, little attention is given to the data used to train these algorithms, and to their impact on the generalization that can be made from these systems. The goal of this talk is to develop, in interaction with the audience, a rigorous methodology for building databases using the example of voice drowsiness detection, and to show that with these, it is possible to design epistemologically robust and meaningful systems.