Our goal is to automate intelligent behavior by building robust probabilistic models for a complex world. The work has a strong basic research component that intersects artificial intelligence, machine learning, computer science, information theory and mathematical statistics. The results of this methodological work are applied to both scientific and industrial applications.