Timeline for What are some well known improvements over textbook MCMC algorithms that people use for bayesian inference?
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 31, 2012 at 16:45 | comment | added | Manoel Galdino | I know almost nothing about ABC, but I would be glad if anyone could explain a bit about the differences between ABC and MCMC methods... | |
| Dec 27, 2012 at 16:47 | comment | added | Rafael S. Calsaverini | Slice sampling is very nice for statistical mechanics models where it's easy to sample uniformly from the set of states which have a given energy. But, as usually in those models energy is a complicated function of discrete variables, it's generally not possible to invert them. | |
| Dec 26, 2012 at 21:54 | comment | added | guy | You may already know this, but slice sampling is pretty easy to implement and avoids some pitfalls of a typical "random-walk" Metropolis algorithm. A problem with traditional Metropolis algorithms is random-walk type behaviour; rather than move purposefully towards good states, they stumble around, moving slowly to the good areas. This is the motivation behind methods that make use of information in the gradient, such as HMC, but slice sampling also falls into this camp and is more automatic. | |
| Dec 26, 2012 at 20:59 | answer | added | David J. Harris | timeline score: 21 | |
| Dec 26, 2012 at 19:02 | history | tweeted | twitter.com/#!/StackStats/status/284011360109613056 | ||
| Dec 26, 2012 at 18:30 | history | edited | Rafael S. Calsaverini | CC BY-SA 3.0 |
edited title as suggested
|
| Dec 26, 2012 at 18:23 | history | edited | gung - Reinstate Monica | CC BY-SA 3.0 |
added tag; formatted; removed thanks; light editing & streamlining
|
| Dec 26, 2012 at 18:20 | comment | added | user10525 | (+1) OK, a nice general-purpose adaptive algorithm "recently" published and already implemented in R, Python and Matlab is the twalk. It works well but it is always a good practice to double-check using another method. BTW, MH is implemented in the R package mcmc. Of course, there are many others but most of them are not implemented in this level of generality and/or they are difficult to implement. Another popular area nowadays is Sequential Monte Carlo. I hope this helps. | |
| Dec 26, 2012 at 18:12 | review | First posts | |||
| Dec 26, 2012 at 18:23 | |||||
| Dec 26, 2012 at 18:08 | comment | added | Rafael S. Calsaverini | @Procrastinator I don't need all of them. Just the most popular and basic. The fact that there are so many is exactly what motivates my question. I can't manage to decide on a good starting point to study the subject because it is vast and seems to have a lot of specialized sub-branches. I'm just asking specialists for a starting point here. :) | |
| Dec 26, 2012 at 18:05 | history | edited | user10525 |
edited tags
|
|
| Dec 26, 2012 at 18:03 | comment | added | user10525 | There are so many that it is probably impossible to list them all here. | |
| Dec 26, 2012 at 18:01 | comment | added | Rafael S. Calsaverini | @StasK, Yes, I'm mainly interested in bayesian models and statistical physics models (which is just bayesian inference on gibbs-like distributions p(x) = 1/Z exp(-E(x)/T) ). Sorry for failing to mention that. | |
| Dec 26, 2012 at 17:59 | comment | added | StasK | Do you mean Markov chain Monte Carlo? The textbook improvements to Monte Carlo simulations that I can think of involve antithetic and/or stratified sampling, as well as quasi-Monte Carlo. Your mentioning of only Gibbs and Metropolis-Hastings is indicative of Bayesian computing, though. | |
| Dec 26, 2012 at 17:55 | history | asked | Rafael S. Calsaverini | CC BY-SA 3.0 |