Dictionary:Simulated annealing

From SEG Wiki
Jump to navigation Jump to search
Other languages:

{{#category_index:S|simulated annealing}} Algorithms based on an analogy between optimization and the growth of long-range order, such as the growth of large crystals in a slowly cooling melt. It is usually implemented by a Monte Carlo ‘‘drunkard’s walk’’ through model space where steps begin in random fashion but progressively become more biased toward stepping in the uphill direction. Simulated annealing has three components: an ‘‘energy function’’ which defines the problem in terms of a parameter set (including interactions between parameters), an ‘‘order function’’ which measures coherence, and a ‘‘temperature’’ which regulates the system’s energy and order (high temperature implies high energy and low order). See Vasudevan et al. (1991) and Sen and Stoffa (1991, 1624–1638).