I am a research consultant at Facebook AI Research. I held research positions at NEC Labs from 2002 to 2014, AT&T Labs from 1996 to 2002, Bell Laboratories from 1989 to 1996. I was with the Institute for Control Sciences in Moscow between 1961 and 1989, eventually holding the position of head of the computer science research department.
I am a member of the National Academy of Engineering since 206, and have held affiliations with the Royal Holloway University and Columbia University.
I am the recipient of the 2005 Gabor Award, the 2008 Paris Kanellakis Award, the 2010 Neural Networks Pioneer Award, the 2012 IEEE Frank Rosenblatt Award, the 2012 Benjamin Franklin Medal in Computer and Cognitive Science, the 2013 C&C Prize from the NEC C&C Foundation, and the 2014 Kampé de Fériet Award.
I have made contributions to the theoretical foundations of machine learning, particularly the notions of capacity and consistency, and I helped develop the Support Vector Machine method.
My current interests include:
1. Rigorous solutions of Statistical Inference problems. I am working on a rigorous formulation (with rigorous solutions) of the problems of conditional probability estimation, conditional density estimation, regression estimation, two densities ratio estimation and so on.
2. Learning with Intelligent Teacher: In most existing machine learning settings, the “teacher” is very simple, merely providing target values. I am working on a model of Intelligent learning which better describes methods of human-students learning than existing classical machine learning models. Along with standard training data the Intelligent Teacher supplies Students with privileged information which include comments, explanations, metaphoric reasoning and so on.. This allow one to include in the methods of learning from data additional learning mechanisms: the mechanisms of knowledge propagation from Teacher to Student. These mechanisms existing machine learning techniques ignore.
3. Statistical Learning Theory: I continue to have interest in the subjects I worked on before: Statistical Learning Theory (the VC theory) that I developed jointly with A. Chervonenkis in the 1970s and in the Support Vector Machine method that I developed in 1990s jointly with my colleagues at AT&T.
Theoretical foundations of learning; rigorous solutions of statistical inference problems; learning with intelligent teachers; knowledge representation and propagation