Efficient model selection in switching linear dynamic systems.

The computation required for a switching Kalman Filter (SKF) increases exponentially with the number of system modes. In this paper, a systematic graph representation for a switching linear dynamic system (SLDS) is proposed along with a \pp{minimum-sum clustering} method to reduce the switching mode cardinality offline, before collecting measurements. It is shown that if mode detection is perfect then the induced mode clustering error can be quantified exactly. The simulation results verify that clustering based on the proposed framework effectively reduces model complexity given uncertain mode detection and that the induced error can be well approximated.