Talk:Rejection sampling

1. as a layman I find the present page too hard to understand, e.g. I don't know the terms "subgraph of a function" or "marginally"
2. how does it compare to alternatives?
3. I was looking for this information using the term "rejection algorithm", is that a common alternative name?
4. Yes, I agree. Too full of jargon to be useful for outsiders! E.g. I have no idea what is an 'instrumental distribution'. In the algorithm, how does one 'sample from g(x)'? —Preceding unsigned comment added by 71.95.156.92 (talk) 15:04, 25 July 2010 (UTC)

Example is good, but incomplete

The circle in square example is very clear, but it lacks the random rejection aspect of the algorithm; once the point is chosen, whether it will be accepted or not is already determined regardless of u. It would be nice to have a more complete example.

Another way of saying this is that this example is more of an example for importance sampling, which does not involve a random u.

—Preceding unsigned comment added by Rodrigo braz (talk • contribs) 02:12, 23 December 2007 (UTC) Rodrigo de Salvo Braz (talk) 02:15, 23 December 2007 (UTC)

Example is AWFUL!

The example reads like someone who completely missed the point and doesn't really understand rejection sampling wrote it. There is an excellent example in rejection sampling from someone who actually understands it at:

http://www.wikicoursenote.com/wiki/Acceptance-Rejection_Sampling

You don't simply define a region of space, select random samples, then accept if they are in that space and reject otherwise. That could also be called "rejection sampling", but has nothing to do with what we're talking about here. I'm deleting the example because it's so God-awful. If someone can come up with an example closer to the one above ,I think that would be very nice. — Preceding unsigned comment added by 149.97.32.36 (talk) 23:12, 19 September 2011 (UTC)

Naming

Is this method perhaps known in English as von Neumann method? And if it is, should we make a redirect (I tried to find it as a von Neumann method, since it is what we call it in Serbian)? -- Obradović Goran (ta^lk 19:16, 28 April 2008 (UTC)

I never heard of it as the von Neumann method and none of my books cite it as such. No objections on the redirect though. —Preceding unsigned comment added by 130.226.87.164 (talk) 14:54, 14 October 2010 (UTC)

Donald Knuth's The Art of Computer Programming Vol. 2 (3rd ed. p125) introduces it as "von Neumann's rejection method". But only "rejection method" is italicized, so I don't think he intended "von Neumann" to be part of the name of the method. Only "rejection method" is in the index. KaiaVintr (talk) 15:39, 23 August 2026 (UTC)

Lacks a graph. Incomprehensible viz indigestible without some illustration, and preferably an example illustration

Needs illustration 77.198.158.225 (talk) 23:34, 26 December 2021 (UTC)

The description is wrong

Item 2 in the description reads "Draw a vertical line at this x-position, up to the maximum y-value of the probability density function of the proposal distribution".

I believe it's a mistake and the word "maximum" should not appear there, so it should read "Draw a vertical line at this x-position, up to the y-value of the probability density function of the proposal distribution". Doobiefletzet (talk) 08:44, 21 December 2023 (UTC)

I thought so too. Deleted.—St.Nerol (talk, contribs) 08:48, 9 April 2025 (UTC)

"Advantages over sampling using naive methods"

The content for this section doesn't seem to match the title, and I'm not sure what it is aiming to explain (also, not sure why inverse transform sampling would be called a "naive" method).

• Refers to inverse transform sampling as a way to "simulate" X, which is confusing to me.
• Much of the text seems to be just a restatement of the rejection sampling method.
• It seems to be talking about why rejection sampling is inefficient, not inverse transform sampling.
• "The Natural Exponential Family (if it exists) [...]" makes no sense, because obviously the natural exponential family of probability distributions exists.

Maybe the intent was to introduce a refinement to the basic method, to handle cases where rejection rate is too high, but it is not explained, and would not make sense under this section title. I can only guess what that other method might be.

I have added "Unreferenced section" and "Confusing section" tags. I will delete the section unless someone else can fix it. KaiaVintr (talk) 18:59, 22 August 2026 (UTC)