Proof — Simulated annealing
Statement
A stochastic optimization method that finds global minima of cost functions with multiple local minima. It emulates the physical process of slowly cooling a solid: at high temperature the system explores widely; as temperature decreases, it converges toward low-energy (low-cost) states. Uses the Metropolis algorithm to accept worse solutions with probability exp[-(ΔE)/T].
Formula
P(accept) = exp[-(ΔE)/T]
Source
Kirkpatrick, Gelatt, Vecchi, Optimization by Simulated Annealing, Science 1983.
Domains
mathematics, computer-science