While studying in Cheboksary, I was conducting some DSP/numerical analysis-related research at Dynamics LTD , designing a fast algorithm for piece-wise differentiable sinusoidal interpolation (see publications). Dr. Sergey Kuzmin and his research group in Dynamics are planning to hardware-implement it in a new model of relay-tomograph testing devices.

Completing the academic requirements of the ChSU one year in advance, I spent a year (98-99) in Moscow, conducting independent research at the Center of Macroeconomic Research and Forecasts in Plekhanov Economic Academy. At the same time I was working as a system programmer in Delphin-Informatika, LTD - a Russian software developer partner of the Energy branch of Siemens AG.

I spent a year (99-00) in Edinburgh doing MSc in non-symbolic Artificial Intelligence (distinction, the best dissertation award) and investigating Products of Experts with Dr. Chris Williams. The model has been recently introduced by Geoffrey Hinton ( UCL / Toronto), as an alternative to mixture-of-experts models for density estimation. Unlike mixture models, PoEs yield sharp posterior distributions in high-dimensional spaces, what makes them potentially efficient for prediction and classification of inherently localized data.

In 00-01 I was leading the discussions of the Probabilistic Modeling group at the Chair of Manufacturing Technology (Erlangen, Germany) addressing engineering applications of probabilistic data analysis and focusing specially on time series classification. I was also an informal co-supervisor of two data-analysis related diploma projects.

In 2002-2005 I was a PhD student working with Dr. David Barber at large scale stochastic systems in the Institute for Adaptive and Neural Computation (part of the School of Informatics, UoE). The principal contribution of the PhD work was a variational approach to maximizing the mutual information in large stochastic environments, and theoretical results relating the method to conditional training in stochastic self-supervised models. I passed my viva on Nov. 11 without corrections (top mark by the UoE regulations).

My current research interests include theory and applications of probabilistic AI, machine learning, and graphical modeling. I am slowly moving to learning more about reinforcement learning in partially observed environments and multi-agent generalizations. The greatest life passions are my family and learning. Some research-irrelevant interests are wild forest hiking and general reading. I am also actively following Russian domestic news and trying to make predictions of the course of the Russian politics and its consequences.

Back to the main page.