A simple approach to time-inhomogeneous dynamics and applications to (fast) simulated annealing
نویسنده
چکیده
We study time-inhomogeneous stochastic dynamical systems with vanishing noise. Conditions implying strong ergodicity are proposed; the asymptotic probability distribution is characterized and we estimate the speed of relaxation. Applying these results to simulated annealing enables us to introduce systems that converge faster than a classical annealing dynamics determined by a Gibbs factor.
منابع مشابه
A simple approach to time { inhomogeneous
We study time{inhomogeneous stochastic dynamical systems with vanishing noise. Conditions implying strong ergodicity are proposed; the asymptotic probability distribution is characterized and we estimate the speed of relaxation. Applying these results to simulated annealing enables us to introduce systems that converge faster than a classical annealing dynamics determined by a Gibbs factor.
متن کاملA Mushy State Simulated Annealing
It is a long time that the Simulated Annealing (SA) procedure has been introduced as a model-free optimization for solving NP-hard problems. Improvements from the standard SA in the recent decade mostly concentrate on combining its original algorithm with some heuristic methods. These modifications are rarely happened to the initial condition selection methods from which the annealing schedules...
متن کاملA Mushy State Simulated Annealing
It is a long time that the Simulated Annealing (SA) procedure is introduced as a model-free optimization for solving NP-hard problems. Improvements from the standard SA in the recent decade mostly concentrate on combining its original algorithm with some heuristic methods. These modifications are rarely happened to the initial condition selection methods from which the annealing schedules start...
متن کاملA Mushy State Simulated Annealing
It is a long time that the Simulated Annealing (SA) procedure is introduced as a model-free optimization for solving NP-hard problems. Improvements from the standard SA in the recent decade mostly concentrate on combining its original algorithm with some heuristic methods. These modifications are rarely happened to the initial condition selection methods from which the annealing schedules start...
متن کاملA Simulated Annealing Algorithm for Unsplittable Capacitated Network Design
The Network Design Problem (NDP) is one of the important problems in combinatorial optimization. Among the network design problems, the Multicommodity Capacitated Network Design (MCND) problem has numerous applications in transportation, logistics, telecommunication, and production systems. The MCND problems with splittable flow variables are NP-hard, which means they require exponential time t...
متن کامل