Faculty of Computer Science Staff Attend International Conference on Machine Learning
On August 6-11 the 34th International Conference on Machine Learning was held in Sydney, Australia. This conference is ranked A* by CORE, and is one of two leading conferences in the field of machine learning. It has been held annually since 2000, and this year, more than 1,000 participants from different countries took part.
The conference focused on the most promising areas of research in machine learning: deep learning, probabilistic models of data processing, scalable optimization, and reinforcement learning. About 500 reports were presented.
The following staff members from the Faculty of Computer Sciences participated in the conference: Arsenii Ashukha, Dmitry Vetrov, Ekaterina Lobacheva, Dmitry Molchanov, Kirill Neklyudov, Evgeny Sokolov and Nadezhda Chirkova.
One report representing Russia, on ‘Variational Sparsifies Dropout Deep Neural Networks’ by Dmitry Molchanov, Arsenii Ashukha and Dmitry Vetrov, staff members from the International Laboratory of Deep Learning and Bayesian Methods, was presented at the conference. It is dedicated to the application of Bayesian methods for the thinning and regularization of neural networks and sparked considerable interest among the researchers. Dmitry Molchanov presented the report at the main track of the conference dedicated to deep learning.
Ekaterina Lobacheva and Nadezhda Chirkova took part in the workshop on natural language processing.
‘In the main track of the conference, papers were presented in nine parallel sessions, unfortunately, we could attend only the sections on our research field (deep learning, recurrent neural networks, language models). However, we were able to discuss other interesting reports at the evening poster sessions. I found the section about recurrent neural networks interesting, as I learned about new neural network architectures, and some interesting results in the field of the exploding and vanishing gradient problem were also presented. We presented our project at the workshop on generating natural language. During the discussion of the article, the participants suggested several ideas for developing the approach further, beyond the research plans that we outlined’ says Nadezhda Chirkova, Research Assistant at the International Laboratory of Deep Learning and Bayesian Methods.