Significance Measure with Nonlinear and Incommensurable Observations

Cross-layer design and optimization normally involve many system parameters, where each parameter may have a different nonlinear relationship with the system performance metric of interest. Furthermore, interdependency among system parameters may significantly affect the effectiveness and efficiency of cross-layer design. However, it is very difficult to derive the interdependency among system parameters and their different nonlinear relationships to the system performance, especially in a dynamic complex networking system, due to uncertainties and randomness existing in data observation and system modeling. In this paper, we propose a new approach for characterizing the complicated interdependencies existing in cross-layer design. The major contributions made in this paper are: 1) nonlinear and incommensurable observation data are characterized by a data preprocessing procedure, and 2) interdependencies among system parameters under uncertainties are then quantitatively measured by using non-additive measure theory. Simulation results show effectiveness and efficiency of the proposed method. Furthermore, the proposed method can be easily implemented and incorporated in existing cross-layer design and optimization.