About the Centre

Recent years have proven that the more data analyzed and the larger the models analyzing it, the better the results. A breakthrough in machine learning occurred due to the successful application of deep over-parameterized neural networks, which proved to be an extremely powerful tool for working with vast amounts of data. However, the theoretical foundations of deep learning still require further research.

As practice has shown, solving complex data analysis problems requires building probabilistic models in which neural networks will be one of the key elements. The standard tool for building such models is the apparatus of Bayesian modeling. A vivid example of the successful synthesis of the Bayesian approach with deep learning are diffusion models, used for generating various types of data.

The Centre was established in 2017 on the basis of the Bayesian Methods group of D.P. Vetrov and conducts research in the paradigm of merging neural network and Bayesian machine learning models: in the fields of deep learning, stochastic optimization, tensor decompositions and model compression, scalable variational inference, diffusion models, etc.

The founder of the Centre is Dmitry Vetrov — PhD in Physical and Mathematical Sciences, Research Professor at the Faculty of Computer Science of HSE University, Professor at Constructor University (Bremen). He is a graduate of the Faculty of Computational Mathematics and Cybernetics of  Moscow State University. At various times, he worked at the Computing Centre of the Russian Academy of Sciences, Moscow State University, the Kurchatov Institute, Skoltech, and Yandex. He is a member of the ELLIS association (European Laboratory for Learning and Intelligent Systems).

The head of the centre is Aibek Alanov, Candidate of Computer Sciences, lecturer at the Faculty of Computer Science of HSE University, and research scientist at the Artificial Intelligence Research Institute (AIRI). At various times, he has worked at Yandex, WorldQuant, and the Samsung AI Centre.