Rao-Blackwellised Particle Filter for Gene Regulation Networks

Abstract

We propose two stage state-space models for genetic networks and estimate the parameter using rao-blakwellised particle filter. This article offers three novel improvement strategies of RBPF. One is that considering multiple samples suits for time-course data (panel data), which seldom appear in engineer application but it was common in biology area. Another is proposing two-stage state space model for gene regulation network. The third is that we modified the RBPF, which requires single kalman filter iteration per input sample. A simple illustrative example and real world SOS data show that our method significantly reducing computational complexity and obtaining good convergence. All of those make the algorithm in this paper possible for real-time implementations.