Probabilistic Numerical Methods for Machine Learning: recent trends


The conference will be held in the lecture hall L001, building L at Faculté de Sciences (from the bus stop IUT, you can follow an avenue that will lead you to the building L, see a map here).

-Keywords : Particle methods, Neural Networks, GAN, Wasserstein distance, Stochastic Optimization and MCMC methods, Robustness and sensitivity.

-Abstract : Probabilistic numerical methods are at the heart of machine learning algorithms. They play an important role in related optimization problems and also certainly for sampling methods which can be used for calibration, generation of new datas or sensitivity problems (among other applications). The increasing complexity of machine learning algorithms involves many new theoretical and practical problems. The workshop thus  aims at focusing on new challenges in probabilistic numerical methods for machine learning, especially in particle methods, stochastic optimization and MCMC, robustness and sensitivity, and Wasserstein computation problems.