Learning population dynamics models from data and domain knowledge

This paper is concerned with integrating knowledge-based modeling or modeling from first principles, with data-driven or automated modeling of dynamic systems. The approach presented here includes methods for equation discovery: unlike mainstream system identification methods, which work under the assumption that the form of the equations is known, equation discovery systems explore a space of possible equation structures. We propose a formalism for representing knowledge about processes in population dynamics domains, and a method to transform such knowledge into an operational form that could be used by equation discovery systems. We also describe the extensions of the equation discovery system Lagramge necessary to incorporate this kind of knowledge in the process of equation discovery.