42
$\begingroup$

I was going to answer a question about reinforcement learning and wanted to show some formulas using the same notation I use on CrossValidated, for instance:

$r_{t+1}+\gamma \max_a Q(s_{t+1},a)$

But it is currently not supported, at least the way I tried it. Can we have support for LaTeX formatting here?

Examples:

$\endgroup$
2
  • 1
    $\begingroup$ You'll need a very good reason / a lot of feature use because LaTeX is very *mumble mumble* expensive *mumble mumble*. $\endgroup$
    – wizzwizz4
    Aug 2, 2016 at 20:08
  • 3
    $\begingroup$ Extremely urgent, please !!! $\endgroup$
    – pasaba por aqui
    Feb 13, 2018 at 16:02

7 Answers 7

Reset to default
22
$\begingroup$

Yes, absolutely. Regardless of your position on the whole "theory vs. implementation" thing, math is an essential part of AI and having convenient access to LaTex would be a boon here.

$\endgroup$
14
$\begingroup$

Here are several questions and answers that would benefit from MathJax support on this website. These are just a few examples I've found in a 5 minutes search. Nevertheless, I think this number is enough to justify a MathJax support on this website.

Questions

Answers

$\endgroup$
7
$\begingroup$

I think we should discourage the use of LaTeX, but should allow it. Our goal is to attract experts in AI, and the language of AI (today) is math. Like that post in the OP (which I wrote, btw), I think math makes a lot of concepts easier to understand.

I think this SE should focus on the design aspects of AI and AI research instead of the programming and libraries (those questions should go to Data Science) or the statistics (those should go to Cross Validated), but some mathematics is often a core component of AI theory.

$\endgroup$
1
  • 9
    $\begingroup$ Please explain why you think we should discourage the use of LaTeX? $\endgroup$
    – Neil Slater
    Jul 15, 2018 at 21:15
5
$\begingroup$

I, for one, would love its inclusion. I do not believe it is possible to divorce AI from mathematics on many levels. For instance, I wanted to ask a question regarding the use of backpropagation with regards to the ANFIS model but had to do so in a clumsy way as I was not able to include the proper notation for partial derivatives. It would surprise me to think that this site is just for "high level" philosophical discussions on AI.

$\endgroup$
4
$\begingroup$

While it might be nice to have for some questions, most questions you would need LaTeX for should be off-topic here. This site is not meant for machine learning questions, as Cross Validated and Data Science Stack Exchange sufficiently cover those subjects.

See: Are all questions asked on stats and data science SE also on topic here?

Note: I posted this answer when I didn't know very much about AI, and I have misunderstood or missed some of the parts of AI that should be on-topic here, I am now of the opinion that we should have LaTeX here. I'll leave this answer here because of the votes (and vote balance) on it, but I don't agree myself anymore with it. So please count an extra downvote from me.

$\endgroup$
5
  • 1
    $\begingroup$ This is a good reason against. A very good reason, in fact. $\endgroup$
    – wizzwizz4
    Aug 2, 2016 at 20:17
  • 6
    $\begingroup$ I disagree @wizzwizz4 . We want to attract experts in AI, and experts in AI speak math. While the questions here might be different from those in crossvalidated and datascience, the audience is expected to have a big overlap $\endgroup$
    – Harsh
    Aug 4, 2016 at 18:21
  • $\begingroup$ @Harsh I agree, which is why I said "it might be nice to have [it] for some questions". $\endgroup$
    – wythagoras
    Aug 4, 2016 at 19:06
  • $\begingroup$ @Harsh Please post an answer with your opinion, so people can vote on it too. $\endgroup$
    – wizzwizz4
    Aug 4, 2016 at 20:37
  • 4
    $\begingroup$ "While it might be nice to have for some questions, most questions you would need LaTeX for should be off-topic here." -- that's just not true. It is not the case that you only need math if you're talking about machine learning. Math is the language of science in general. If we want to talk real, serious science here, we should have MathJax here. $\endgroup$
    – mindcrime Mod
    Aug 18, 2016 at 15:35
4
$\begingroup$

I would like to add to the calls for LaTeX support with a specific topic.

In my opinion, the AI Stack Exchange should be the home for questions about Reinforcement Learning.

RL questions actually appear in larger numbers on Data Science and Cross Validated Stack Exchange sites. That makes little sense to me, when AI, robotics and other better homes in a conceptual sense exist for this topic.

RL is a technical subject requiring solid understanding of underlying maths, especially for anyone wanting to engage in algorithm design. I would like to be able to write equations and maths-based pseudo code when writing questions or answers about RL. It is a shame that this site presents a barrier to doing that. Along with the larger audience for other Stack Exchange sites, this one is losing out IMO on a current hot topic that could provide much traffic. And in part that is due to barriers when writing content.

$\endgroup$
1
  • 1
    $\begingroup$ You're the man Neil! $\endgroup$
    – DukeZhou
    Aug 13, 2018 at 21:01
4
$\begingroup$

Starting a new list of math questions to expand on nbro's list:

$\endgroup$
2
  • 1
    $\begingroup$ can add ai.stackexchange.com/q/6308/1641 to the list. The question would be a lot more clear if the Equation that it's about could be comfortably written in the question, rather than requiring people to download the linked paper and find the Equation. My answer to the question would also be much more readable with proper MathJax over... whatever I ended up doing. $\endgroup$
    – Dennis Soemers Mod
    Aug 7, 2018 at 13:13
  • $\begingroup$ @DennisSoemers Added. (Feel free to directly update this list. Nbro's list original list, I think, is static at this point, so I created this one as a supplement.) $\endgroup$
    – DukeZhou
    Aug 7, 2018 at 19:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .