Inverse Problems in Complex Multi-Modal Regulatory Networks Based on Uncertain Clustered Data

Complex regulatory networks effected by noise and data uncertainty occur in many OR applications. The complexity is compounded by the unknown interactions between the system variables that have to be revealed from unprecise measurement data. The concept of target-environment networks provides a generic framework for the analysis of complex regulatory systems under uncertainty. Data mining methods like clustering and classification can be applied for an identification of functionally related groups of targets and environmental factors. The effects of the intricate connections between target and environmental clusters on single entities are determined by a parameterized time-discrete model. A crisp regression problem is introduced for parameter estimation and in case of uncertain data, ellipsoids are used to describe the clusters and error sets what refers to particular robust counterpart programs.