Establishing an acceptable model of an observed dynamic system from observations is a challenging task that occupies a major portion of the mathematical modeler's effort. In this talk, I address the task of inducing models of dynamic systems from data and background knowledge. The models are cast in terms of processes that govern the behavior of the system. I present an approach to inducing process models that extends the existing approaches in two directions. First, the method relies on an extended formalism that organizes process models in a hierarchical manner. Second, the approach utilizes heuristic rather than exhaustive search to explore the space of process models. I illustrate the method's operation on examples of environmental and biochemical systems from time-series data.

The work presented in the talk was done in collaboration with Pat Langley, Oren Shiran, and Will Bridewell.