Nonstationary Markov chains and convergence of the annealing algorithm (Q1101782): Difference between revisions

We study the asymptotic behavior as time \(t\to +\infty\) of certain nonstationary Markov chains, and prove the convergence of the annealing algorithm in Monte Carlo simulations. We find that in the limit \(t\to +\infty\), a nonstationary Markov chain may exhibit ``phase transitions''. Nonstationary Markov chains in general, and the annealing algorithm in particular, lead to biased estimators for the expectation values of the process. We compute the leading terms in the bias and the variance of the sample- means estimator. We find that the annealing algorithm converges, if the temperature T(t) goes to zero no faster than \(C/\log (t/t_ 0)\) as \(t\to +\infty\), with a computable constant C and \(t_ 0\) the initial time. The bias and the variance of the sample-means estimator in the annealing algorithm go to zero like \(O(t^{-1+\epsilon})\) for some \(0\leq \epsilon <1\), with \(\epsilon =0\) only in very special circumstances. Our results concerning the convergence of the annealing algorithm, and the rate of convergence to zero of the bias and the variance of the sample-means estimator, provide a rigorous procedure for choosing the optis, from those presented at the Congress mentioned in the title. The volume is divided into five sections. Section I is on complex system modelling, simulation and identification (7 papers). Environmental system, production engineering, turboset control systems and mechanical vibration test benches are considered. Section II is on bond graph analysis and modelling (15 papers). Bond graph theory is applied to rigid body systems, a fluid filled pipe, unsteady state heat conduction, an adaptive vehicle air suspension, a feel force system, a dynamic robotic model and manipulator control systems. Section III is on nonlinear oscillators and chaotic systems (6 papers). Section IV is on distributed parameter systems (13 papers). Stability analysis, identification by Walsh functions, parameter and state identification, adaptive state estimation and computational methods for an optimal control problem are discussed. A literature overview of the last two decades is given. Section V is on control of complex systems (8 papers). This volume represents five major areas of system theory, and shows the modern trend in the growing field of complex and distributed parameter system theory.