Diffusions for global optimizations
暂无分享,去创建一个
We seek a global minimum of $U:[0,1]^n \to R$. The solution to $({d / {dt}})x_t = - \nabla U(x_t )$ will find local minima. The solution to $dx_t = - \nabla U(x_t )dt + \sqrt {2T} dw_t $, where w is standard (n-dimensional) Brownian motion and the boundaries are reflecting, will concentrate near the global minima of U, at least when “temperature” T is small: the equilibrium distribution for $x_t $, is Gibbs with density $\pi _T (x)\alpha \exp \{ - {{U(x)} / T}\} $. This suggests setting $T = T(t) \downarrow 0$, to find the global minima of U. We give conditions on $U(x)$ and $T(t)$ such that the solution to $dx_t = - \nabla U(x_t )dt + \sqrt {2T} dw_t $ converges weakly to a distribution concentrated on the global minima of U.
[1] N. Metropolis,et al. Equation of State Calculations by Fast Computing Machines , 1953, Resonance.
[2] Donald Geman,et al. Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images , 1984, IEEE Transactions on Pattern Analysis and Machine Intelligence.
[3] B. Gidas. Nonstationary Markov chains and convergence of the annealing algorithm , 1985 .