#  The Future of Bayes 

 





November 18, 2010

 

 

John Salvatier has a blog post on the [future of MCMC algorithms](http://goodmorningeconomics.wordpress.com/2010/11/16/the-promise-of-bayesian-statistics-pt-2/), focusing on differential methods, which use derivatives of the posterior to inform where the algorithm should move next. This allows for greater step length, faster convergence, and better handling of multimodal posteriors. Gelman [agrees](http://www.stat.columbia.edu/~cook/movabletype/archives/2010/11/derivative-base.html) with the direction. There has been some recent work on implementing [automatic differentiation in R](http://gsoc2010r.wordpress.com/2010/08/20/automatic-differentiation-in-r/), which is the cornerstone of the algorithms Salvatier discusses. Perhaps we will see this moving into some of the more popular MCMC packages soon.

On a slightly different Bayes front, SSS-pal and former blogger [Justin Grimmer](http://www.justingrimmer.org/) has a [paper](http://pan.oxfordjournals.org/content/early/2010/11/11/pan.mpq027) on variational approximation, which is a method for deterministically approximating posteriors. This approach is often useful when MCMC is extremely slow or impossible, since convergence under VA is both fast and guaranteed.

Posted by [Matt Blackwell](http://www.iq.harvard.edu/blog/sss/archives/author/matt-blackwell/) at November 18, 2010 9:52 AM



 

 

 



 

 

 Share on:- [     Facebook ](#)
- [     Twitter ](#)
- [     Linkedin ](#)