Jump to content

Talk:Parallel tempering

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Untitled

[edit]

Correctly attributed the technique to Swendsen, and added formula.

Not sure about the other credits, but left them in.

Terminology challenge

[edit]

I feel that the use of the word 'temperature' in this article is confusing. The article presents a technique for improving the fit to multimodal target distributions in Metropolis-Hastings analysis. Statistics has perfectly good familiar terminology for describing the properties of distributions, such as standard deviation. Further confusion is created by a completely different meaning of temperature in the closely related technique of Simulated annealing. This analogy to metallurgical processes is quaint, but has been stretched far beyond its usefulness for the purpose of imparting knowledge to the technical non-specialist. Doug (talk) 17:15, 5 June 2015 (UTC)[reply]

Further, a probability target is far more general and useful than an energy target distribution. Doug (talk) 17:19, 5 June 2015 (UTC)[reply]