Talk:Bayesian programming

This idea didn't originate with Jaynes

A good start. Jaynes is a prominent Bayesian, although there are many of equal stature. It doesn't make sense to make him so prominent.

I know Jaynes, but not this topic. How does it differ from probabilistic programming (which I also no a bit about..)? Is a separate article really needed? — Preceding unsigned comment added by 155.143.31.206 (talk) 10:26, 4 August 2016 (UTC)

Article is written as a personal opinion or reflection essay

I wrote the first version of this article as a starting basis.

I hope it summarizes the work and opinion of a community of people using "bayesian programming" as one of the possible approach in probabilistic modeling.

My purpose is that this community will contribute in the coming days to improve this article and make it more consensual and less "personal"

I thought it is an adequate starting point but I might be wrong as it is my first "long" contribution to wikipedia.

Erreip (talk) 14:08, 24 November 2013 (UTC)

Article needs additional citation

It is not clear for me at this point where this additional citations are needed.

• The introduction has citations and numerous links to other wikipedia articles.
• The formalism section has no citation but is suppose to be self sufficient as a formal presentation.
• The example section has no citation but refers to other wikipedia articles that provide the required citation.
• The applications and discussion section have numerous citations.

Could you detailed where you think additional citations are required. It would be a great help.

Erreip (talk) 14:08, 24 November 2013 (UTC)

There is no such thing as "self sufficient" on Wikipedia. Everything that is not blatantly obvious (such as "water is wet") must to be sourced to reliable secondary sources. None of those formulas are in the same league of obviousness as 1+1=2. Other wikipedia articles are never acceptable sources because they in turn may also have sourcing problems. College-level textbooks are probably the type of sources this article needs. You must always assumethat your reader knows very little or even nothing at all about the subject. You can get specialist help at WT:WikiProject Statistics, I know less than nothing at all about this stuff. Roger (Dodger67) (talk) 16:35, 25 November 2013 (UTC)

Article is an orphan

I agree that at that point this article is orphan.

We will work on inserting some links to this article at some appropriate places in the coming days.

Erreip (talk) 14:08, 24 November 2013 (UTC)

The other articles about "Bayesian XXXXXX" are ideal places to add links to this article - get advice from WikiProject Statistics. Roger (Dodger67) (talk) 15:40, 26 November 2013 (UTC)

Miscellaneous changes

Notes section

Dodge67 remarks about the section Notes and discussion looking too academic is right.

However, the article needs a separation (new section) to separate the "application" section from all the notes and references.

90.2.245.105 (talk) 22:02, 24 November 2013 (UTC)

The problem is the word "Notes" - on Wikipedia it has a specific use - as a heading for a footnotes section, the content current under it are not footnotes, they look to me like various "explanations". As I said elsewhere, ask WikiProject Statistics for further help. Roger (Dodger67) (talk) 16:42, 25 November 2013 (UTC)

Section "bayesian inference engine"

I agree with Dodge67 that the first version of this section may have been inappropriate.

However, I think that such a section may have its place and interest in this article.

Don't you ?

90.2.245.105 (talk) 22:02, 24 November 2013 (UTC)

Independent refs?

Reading over the article, it looks like a good start. But I have concerns about independent sourcing and possible conflict of interest. 14 of the 27 sources have Bessière (there are two, J. and P., possibly related). At first glance, none of the non-Bessière references seem to be about Bayesian programming in particular. Are there sources on Bayesian programming independent of the research group involving Bessière? The reason this is important is that articles need to be based on multiple independent reliable sources, per WP:RS, for neutrality and notability of the topic, per WP:Notability. If is topic is not notable, the article covering it could be subject to deletion.

The second point is that P. Bessière is Pierre Bessière, whose name looks related to the article creator. If so, then Mr. Bessière has conflict of interest in editing this article and in particular sourcing primarily to papers on which he is an author. See Wikipedia's conflict of interest guidelines at WP:COI for details. It is best to declare a conflict of interest and allow other editors to look over the work and check neutrality. --Mark viking (talk) 20:05, 25 November 2013 (UTC)

Thanks Mark for these remarks.

I (Pierre Bessière) am the writer of the initial version of this article.

May be I did not understand the rules of wikipedia and the article should be deleted.

Here is why I thought it could be legitimate and interesting to write such an article:

• Concerning multiple independent reliable sources and notability: I thought that several tenth of papers in peer reviewed international journals :in very different fields and more than twentieth PhD theses in various universities around Europe could be sufficient.
• Concerning conflict of interest: I have absolutely no interest but a scientific one. If when you write about a subject that you know there is a :conflict of interest, then there is one.

My only interest is to promote and discuss a scientific opinion shared by a community of several tenth of scientists around Europe and south America.

I would be pleased and interested to discuss and argue on any related scientific matter to make this article more interesting.

I am not qualify to discuss wikipedia rules and I am very serious when I said that the paper should be deleted if it infringes these rules.

Erreip (talk) 23:16, 25 November 2013 (UTC)

I am not sure to really understand what you mean by "It is best to declare a conflict of interest and allow other editors to look over the work and check neutrality".

I would be very pleased to do that but I do not really know how to do it.

Indeed, as you may see in history I originally posted the article as "Article for creation/Bayesian programming" and it was moved quickly to "created"

Erreip (talk) 23:41, 25 November 2013 (UTC)

Full disclaimer: I am a previous PhD student of Pierre Bessière (Erreip). I want to provide some input: most of the work done in the framework of Bayesian programming was done with some relation to P. Bessière's research teams (over the years, at INRIA and now at Collège de France). Still, all the publications were peer-reviewed, and they were the product of real-world software that gave significant results (in all areas). P. Bessière was the principal investigator of the foundations of the framework of Bayesian Programming, but it does not disqualify neither the scientific content nor the author himself.

Somehow, this article could be compared to the ones on Hierarchical temporal memory or PLT Scheme for instance... I understand that it should fit the context of other articles better (incoming and outgoing links to other articles), but as far as I'm concerned, when I want to read a book about C, I take the K&R. You may want to take the article offline while it is not up to Wikipedia's standard on this point, but the fact that this is a focused research domain should not be disqualifying by itself!

Last but not least, that is not because no team at Stanford or MIT worked on "Bayesian Programming" per se that it was not (and keep on being) influential in the world of probabilistic programming (as much as Bayesian Programming itself stands on the shoulders of giants), and it has been active in this research area in the past (e.g. NIPS 2008 workshop on probabilistic programming).

SnippyHolloW (talk) 16:25, 26 November 2013 (UTC)

Hey folks; don't worry too much, I don't think that deletion is being discussed right now. There's anyways a long administrative process involved to discuss with the community, wether or not a page should be deleted. My point being: this Talk page is not the place to discuss about the admissibility of the article. There's no public to convince. Not just yet.

A small meta point before moving on: Have a look at Wikipedia:Talk page formatting for the "expected" formatting of talk pages. I indented manually your previous comments here. Good threading really helps in long discussions :)

Mark has a fair point about the conflict of interest, and the diversity of the sources. I know you guys, and I have some notion of the fact that your are not crooks ( ;) ), but the Internet and the Wikipedia community as a whole is not aware of this. While editing this article, be aware of the fact that most of the readers and community members are not specialists, and will have a hard time cross-checking the facts that you advance. (For instance, if Snippy tells me that conference XXX has great content while conference YYY is not a serious place, I'll have a hard time finding by myself if he's stating the truth).

If you can, a good starting point would be to diversify as much as possible sources. If you have papers from other authors, do add them to the references, even if it inflates the number of notes. The point of this exercice is to show to a neutral, ignorant third party, that this article is proper, independent content. Right now, it's true that the number of papers mentioning Pierre are overwhelming; and third parties reading this article will legitimately wonder about the relevance of sources.

NicDumZ ~ 17:22, 26 November 2013 (UTC)

Most of the article content is unsourced - From after the lead right up to the "Applications" sction there is only one very lonely looking reference. As an "ignorant third party" I expect, as a bare minimum, that every one of the mathematical formulas would have at least one reference. It may be perfectly obvious and "self evident" to someone with a PhD in the subject but to someone who barely scraped through high school algebra thirty years ago, it might as well be written in Klingon. The only way I (as an exemplar of the "ignorant third party") can be assured that it isn't utter nonsense is if it has reputable sources. Roger (Dodger67) (talk) 20:05, 26 November 2013 (UTC)

Roger (Dodger67), I sincerely apologize about the "ignorant third party" formulation. It does not sound how I wanted it to be, it was definitely not meant as a personal attack. For the record, I do not personally understand a lot of the content in this article, and you are right that it needs some work. I'm trying to bridge the gap between new editors and the community here, sorry if I sounded unconstructive. NicDumZ ~ 07:31, 27 November 2013 (UTC)

You have nothing to apologize for. The description "ignorant third party" is a completely accurate description of readers such as myself - as I mentioned above, I have no prior knowledge of the subject. Roger (Dodger67) (talk) 11:31, 28 November 2013 (UTC)

Thanks SnippyHolloW and NicDumZ for weighing in and providing context regarding this topic. NicDumZ, you are correct, I am not proposing to delete this article--in fact, it is a worthy topic and a good start on an article. I apologize, Erreip, if I came across as harsh or threatening. You haven't infringed on any rules and I know you and the others are all acting in good faith here. This is more a case of climbing a steep learning curve at Wikipedia. Thanks for declaring your potential conflicts of interest; this will help outside editors such as myself asses the content and contributions to the article.

NicDumZ is correct that the best way to make this article more neutral and implicitly demonstrate notability is to add sources/references from researchers independent of Bessière. Independence isn't a black and white concept, especially in an academic specialty where everyone knows everyone else. Typically, researchers or research groups at other universities or institutions can be considered independent. At Google Scholar, the search term "Bayesian Programming" -Bessière -Bessiere yields about 200 hits. I cannot tell at a glance which of these are using Bayesian programming as a generic term for computing with Bayesian statistics and which are about this topic in particular, but perhaps domain experts like yourselves can find some potential independent refs there. The NIPS workshop also sounds promising. I know the NIPS conference papers are peer reviewed--is this true of the workshop papers as well? The one NIPS workshop I gave a presentation at, was definitely not peer reviewed.

For potentially new editors, when Roger says the article is unsourced, he is referring to the lack of inline citations. While there are many general refs, it is difficult or impossible to tell from which source did a sentence or equation originate. Different editors have different ideas on the number of inline citations needed, see WP:SCICITE for scientific citation guidelines. But almost all editors want to see each section cited and many want at least each paragraph cited. --Mark viking (talk) 21:02, 26 November 2013 (UTC)