Understanding Complex Systems via Data Dependencies

Val Bykovsky, Air Force Research Lab, Hanscom

Real world is being explored via experiments, and a result of any experiment is a dependence between the experiment context and the measurement data. The dependencies are the relationships that depend on the measurement precision only, a kind of basic units of exploration. For complex systems, a direct dependencies analysis complements an equation-based one as it has its limits, and handling by a human expert hundreds of dependencies is a huge challenge. Due to complexity, analysis of dependencies in cell biology, neurodynamics, and similar problems cannot be done with equations. Computer power introduces a new element in the existing exploration paradigm - handling the *complex dependencies* can be done *programmatically*, including generalization of dependencies - thus removing the human limitations. The mapping can be built using statistical, neural and other methods within a data-driven framework that simulates a physical phase space. Then estimates (predictions) can be done using search in such a framework driven by measurement data.