Situation dependant evaluation of regression-type signal processing problems

Regression-type algorithms are widely used for system modeling and characterization. There are applications where such characterizations are to be performed “on-line” to support control mechanisms and other decisions. In embedded autonomous systems robustness considerations ask for techniques, which, in addition to reflecting the actual state of the system and its environment, can continuously provide immediate signal processing results even in case of abrupt changes and/or temporal shortage of computational power and/or loss of some data. There is a need for robust techniques called “situation dependant” or “anytime” algorithms, which can provide short response time and be very flexible with respect to the available input information and computational power. The paper presents some considerations concerning such flexibility in the case of regression-type algorithms.