I just wanted to try and see if I can cook up some simple Python routines to allow translation of commonly used models into code that we can then use some inference engine on. So, basically I am trying my hand at designing some probabilistic programming framework and learning some things in the process.

I wanted to seek some advice on designing some software that can be flexible enough to allow translation of most commonly used graphical models to code. In particular, I wanted some advice on class design and suitable representation (factor graphs versus different representation for directed/undirected graphs etc.). Would this kind of question be considered on-topic on CV? The design issue clearly belongs on SO but the folks here are more knowledgeable regarding the models themselves and can advise better on what these objects should include/represent.

| |
  • 3
    $\begingroup$ Standard advice applies: if it a question needs statistical expertise to understand or answer, ask it here; if it's about an algorithm, routine data processing, or details of the language, then please refer to the collection of links to resources we maintain. By the way, I am not clear that broad questions on software design go down well on SO, but it runs as a set of largely self-contained communities, so who can be certain? The Python users on SO would soon tell you if they thought a question off-topic, as we would here. I fear that your question could seem too broad in any forum, however. $\endgroup$ – Nick Cox Jul 22 '15 at 10:51
  • $\begingroup$ Yeah, I am struggling with that as well. I will try and narrow the scope of my question. The problem is that it requires a lot of domain specific knowledge which is in abundance here :-) $\endgroup$ – Luca Jul 22 '15 at 17:11
  • $\begingroup$ @NickCox this would appear to be a complete answer. Perhaps you should consider positing it as such? $\endgroup$ – Sycorax Jul 22 '15 at 21:11
  • 2
    $\begingroup$ Not that there are a number of computing/programming-related groups other than SO which you might consider; in this case "conceptual questions about software development" go on programmers.SE, so it may be worth considering. See also here: meta.stackexchange.com/questions/129598/… $\endgroup$ – Glen_b Jul 22 '15 at 23:47
  • 1
    $\begingroup$ My personal opinion is that we need to be clear about our content borders. I refer to the variance of the border. It is rare in "sufficiently complex" (in a Godel sense) topics to have a clear border. Given the power of computing and its place in statistics the border must be turbulent, evolving in time, and hard to measure. A value-creating rule, one that improves and sustains community health and relevance, is going to account for that. $\endgroup$ – EngrStudent Sep 21 '15 at 17:05

You must log in to answer this question.

Browse other questions tagged .