Wolfram Mathematica

2:52 AM

@DanielLichtblau Saw your comment - delete at leisure. Thanks for the answer!

6:48 AM

what happened to the chat link at the mathematica.stackexchange.com homepage? Also, for tracked symbols :> behaves the same as ->, can I just use -> and not worry about it?

Mr.Wizard

7:03 AM

@AdamDreaver I still see a "chat" link at the top of the page in the tool bar area. Is it missing in your browser?

Adam Dreaver

@Mr.Wizard I didn't realize that was for this room, I thought that was just a general chat link; I'm referring to a link to this chatroom which was on the right side bar of the webpage.

Mr.Wizard

@AdamDreaver The link at the top should take you to a page that shows chat rooms that are specifically linked to Mathematica.SE (there can be more than one). I cannot confirm that another direct link was removed as I didn't make use of it. You can ask or search on the main Meta.StackOverflow site if it troubles you.

@AdamDreaver It does seem you can use -> for TrackedSymbols in many cases but it may be more robust to use :>. I think I would prefer the latter merely as a matter of principle.

Adam Dreaver

why is :> more robust than -> , my only analogy is := vs = for functions, but then it makes sense for a delayed assignment on functions.

Mr.Wizard

@AdamDreaver I don't have an example at hand, but since -> does not hold its arguments its theoretically possible for a symbol with a global assignment to evaluate, giving you something like: TrackedSymbols -> {7}. Manipulate has HoldAll and handles this properly either way, but for example Refresh has only HoldFirst and I could probably cobble together a place where this incorrect evaluation takes place.

rm -rf

@AdamDreaver The link in the right sidebar is cycled along with other ads, so you see it only some of the time.

Mr.Wizard

7:18 AM

@rm-rf Thanks. I never paid attention to that.

Adam Dreaver

@Mr.Wizard ty you for the explanation, I'm still an newbie, so it's a little subtle the differences, but I glean your answer.

@rm-rf ty

rm -rf

@AdamDreaver The docs for TrackedSymbols has an example using -> showing how it fails

Mr.Wizard

@rm-rf The documentation you say? Imagine that. :-)

Adam Dreaver

Lol, I already read the doc before I came here, I guess I'm too stupid to not miss something

rm -rf

heh

Adam Dreaver

7:21 AM

ARrrgh

Oh, I didn't click on the possible issues tab, that small text, I'll blame my eyesite

Mr.Wizard

@rm-rf They added that "possible issues" section some time after version 7. Glad it's there now, as a simple case like that didn't come readily to mind.

rm -rf

I spent quite some time today with a complicated (for me) dynamic... feels good to finally have it working

So yeah, I was all inside of related functions' doc pages, so remembered it...

Mr.Wizard

@AdamDreaver As a rule you need to carefully expand all sections of the documentation page to avoid missing important details. They are often fairly well hidden. At the very least look at "Details" (used to be "More information") and "Possible Issues" (if it exists) sections.

Adam Dreaver

@Mr.Wizard Good advice, ty.

rm -rf

@Mr.Wizard Hmm... btw, do you have v9 now?

Mr.Wizard

7:25 AM

Heh... not yet. :^)

Yves Klett

Oh dear...

0

Q: plz can any one solve following question

solve {integeraton symbol(limits from 25 to 50)integration symbol (limits from -5 to 5) Fxy(σ) } where Fxy(σ) is, Fxy(σ)=(1/(2πσxσy))e-((x^2/2σx^2)+(y^2/2σy^2))

integral-equation

Is this a hoax?

@rm-rf ah thanks for that

rm -rf

@YvesKlett :)

Adam Dreaver

Good night all ty for the help

Yves Klett

might have been a contender for worst Q ever!

rm -rf

> integeraton symbol(limits from 25 to 50)

2

LOL

Yves Klett

7:31 AM

@rm-rf plz!

although giving 42 as the answer would also have been very satisfying.

acl

11:07 AM

nice

Dominic Michaelis

12:30 PM

hi

Mr.Wizard

1:30 PM

Hello @Dominic -- sorry nobody was around to greet you.

Dominic Michaelis

oh thats ok

i love futurama :D

Mr.Wizard

:D

Leonid Shifrin

@Mr.Wizard Hello! What's up?

Mr.Wizard

@Leonid I was commenting to Rojo that you should put extra space around ~infix~ operators, and it occurred to me that automation for this would be nice. You are surely the expert on parsing etc. around here, so you should be able to tell me what is and isn't possible.

(1) Is there any way to affect the automatic spacing of code in the FrontEnd?

(2) Is there a way to add spacing to text that is copied as InputText in a way that is parsing-aware?

I would guess "no" to both, but you've already done things I thought impossible.

Leonid Shifrin

@Mr.Wizard I could add this to my formatter. I would not know hoe to do this other than process the box expression. Perhaps, for this particular case the rule would be much simpler.Give me a second...

Mr.Wizard

1:44 PM

Sure.

Leonid Shifrin

@Mr.Wizard Ok, here is the rule: RowBox[{left___, p : (PatternSequence[, "~"] ..), right__}] :> RowBox[{left, " ", Sequence @@ Riffle[p, " "], right}]. You have to extract the boxes from your input cell, subject that boxed expression to this rule, and then use CellPrint to print a new code cell (e.g. using a function such as this: prn = CellPrint[Cell[BoxData[#], "Input"]] &). This can be automated via a palette. I don't see another way of doing this.

@Mr.Wizard Now, this may be not the most general solution, but it can be a start.

Mr.Wizard

@LeonidShifrin Okay, I'll try that out. Thank you. Do you think that such customizations could be written as a // wizardForm wrapper or do you foresee a problem with that?

Leonid Shifrin

@Mr.Wizard Probably fine. Make wizardForm HoldAll, and it should then find the location of the cell being executed, extract the code from it, strip itself, perform this rule application, and insert a new cell in the place of the old one.

Mr.Wizard

@LeonidShifrin lol -- that sounds complicated. I was thinking of something with MakeBoxes.

Anyway, I'll toy with it soon and if I cannot figure it out I'll post a question.

Leonid Shifrin

1:59 PM

@Mr.Wizard Won't work: infix is a box form,MakeBoxes kills it when parsing, since it does not care which form the function has been applied. Been there, done that.

Mr.Wizard

Ah, yes, that's the kind of experience I needed. No point in banging my head on an immovable wall.

The palette sounds like the right approach anyway.

Leonid Shifrin

@Mr.Wizard Actually, a more correct statement is that infix is killed by the parser, before MakeBoxes even looks at an expression

Mr.Wizard

@LeonidShifrin There's always CellEvaluatorFunction, right?

Leonid Shifrin

@Mr.Wizard Yes, but I don't remember if there is a way to extract boxes from it, or alternatively parse a string to boxes only. And if you have to go through parsing to an expression, then this isn't good enough for your purposes.

@Mr.Wizard But it is fairly easy to do it the way I outlined.

Mr.Wizard

I'm pretty sure the input to CellEvaluatorFunction (from Input cells) is the box form.

Leonid Shifrin

2:03 PM

@Mr.Wizard Well, then good for you :-)

Mr.Wizard

Not the raw box form. That takes UndocumentedTestFEParserPacket.

Leonid Shifrin

@Mr.Wizard But I would still do it by extracting boxes using things like NotebookSelection, NotebookRead etc. G.t.g now, but we can discuss later if you wish.

Mr.Wizard

See you later, Leonid.

Leonid Shifrin

See you, @Mr.Wizard.

m_goldberg

2:21 PM

@Mr.Wizard. Concerning my "obsession" with editing -- I prefer to think of it as "dedication" :) -- were you serious in your comment about it earlier? Do you think I being too heavy-handed and should be doing less?

rm -rf

3:33 PM

@m_goldberg I'm sure Mr.Wizard meant that in jest (esp. given the number of glaring mistakes in the OP that he did not fix :P). I'll tell you from experience though, that there will always be someone who was unhappy that their post was edited. I used to edit and fix posts like crazy last year until the site graduated and I've occasionally gotten comments subtly hinting that it would've been better if I had left their post in pristine condition, with all its flaws intact...

In such cases, I just point them to this section in the FAQ

Yves Klett

@rm-rf for new users of SE (like me a year ago) the first edit by another person may seem strange

rm -rf

Yeah, I fully understand that.

m_goldberg

@rm-rf, I suspect you're right about it being a jest, but I wanted to be sure. I worry a lot about being too heavy-handed. Also, I have trouble discerning wheter @Mr.Wizard is joking sometimes. In comments, a few OPs have thanked me for their edits and a few others have complained. Most say nothing, which is what I prefer.

Mr.Wizard

@m_goldberg I meant it lightheartedly, but not in jest. Please understand that I do not think the edit was bad, unwelcome, heavy-handed, or anything of the kind. I simply wouldn't (and didn't) bother myself at this point. (The question was understandable, though hardly brilliant.) I'm just impressed by your endurance, that's all.

As to other times I may be joking: (1) I'm usually lighthearted, but sometimes the grump-pants do come out. (2) It never hurts to ask.

rm -rf

3:51 PM

@Mr.Wizard Not surprising for a 170 yr old professor... ;)

m_goldberg

@Mr.Wizard. I'm glad to learn you didn't mean it to be critical. And I did ask in counter comment. As for my endurance: I'm a retired, old geezer with lots of spare time. and I enjoy the challenge of editing a really bad post.

BTW will there be fake posts on April 1? Did it happen last year?

rm -rf

Not fake posts, but SE does some sort of network wide silliness...

Last year it was the MS Office Clippy that would randomly pop up and nag you. Two years ago, it was unicorns dancing on your screen when you upvote and so on...

m_goldberg

@rm-rf. It seems to me that sort of joke must come from one or people with very high level of privilege or from hackers. Which was it?

rm -rf

SE devs :)

m_goldberg

@rm-rf. Oh, so it was both. :)

I'm off to get lunch. I love MMA.SE not less, but food more.

rm -rf

8:55 PM

@gracenote Since you're on a similar discussion in the TL, I thought I'd ping you on ours as well: meta.mathematica.stackexchange.com/q/744/5

Grace Note

Nice nice nice nice nice nice~

rm -rf

We've had this running for 5-6 mo. now, had several discussions in chat, frequent reminders, a community ad, etc., etc. and set a final deadline of mar. 10. So that's what we have now

Grace Note

Why is it locked, though?

Oh, okay. So, basically, this meta thread is the completed list?

rm -rf

We had to set a deadline for voting, else it kept dragging. We had decided on 10 votes ("Nice Answer") to make it past the goal post (those were the rules laid out in the question)

Grace Note

Mmkay. Have you run this list by the Stack Overflow folks?

rm -rf

8:57 PM

So the list is what the community has vetted and voted on.

@GraceNote To be brutally honest with you, almost 100% of users who were in the SO mathematica tag are here and not active there since last Jan. They are the folks who wrote this stuff and want it here to curate, integrate it with other answers, etc. I honestly couldn't care less what a C# user has to say on these questions...

I don't mean to sound nasty about it, but I do know what it is like to open this discussion on metaso.

Grace Note

Mmkay.

rm -rf

That said, I also don't want to rob them of a voice (although, I adamantly maintain that SO's "community" lives in the tags, and ours has spoken), so I leave that decision to you.

@GraceNote I've discussed this w/ Anna in the past (around the time when we started this discussion) and you can see those bookmarked. If you want more evidence of chat reminders/discussions/etc., let me know and I'll dig it up.

Grace Note

Aye, yours was one of the major cases behind the development of the system, as it were.

halirutan

@GraceNote I just read here blog.stackoverflow.com/2011/08/… about you. This is kind of confusing:

rm -rf

Heh, I know exactly what's going to be quoted now :P

halirutan

9:11 PM

"Chris is a gamer by heart and heritage, and thus a lot of her life comes from his love for video games. "

@rm-rf I though maybe I just don't get it as non-native speaker..

@GraceNote Ah OK, there are more her's in the text. Sorry.

Oleksandr R.

@halirutan en.wikipedia.org/wiki/Genderqueer [This may or may not apply to Grace Note; it really isn't any of my business. But just to let you know that gender is not limited to "man" or "woman".]

rm -rf

@halirutan and there are more his in the text :)

halirutan

Hmm, now I'm really confused and I have the feeling it's non of my business. Sorry again. I don't have to know everything.

@rm-rf Yes, 43:17 for His,his,He,he..

@rm-rf So this confusion is on purpose, right?

Grace Note

Yes, it's deliberate.

rm -rf

@GraceNote here's a convenient list of 10+ voted questions from the poll.

Grace Note

9:22 PM

Thanks

halirutan

@OleksandrR. It seems you are always reading and just jump in when something catches your eye. Amazing ;-)

Oleksandr R.

@GraceNote yes, we know, and we're very grateful! Thank you! I hope that this system was developed at least partly with us in mind won't upset anyone on SO in regard to our possibly being seen to receive special treatment

@halirutan not always but often. :)

Oleksandr R.

10:34 PM

I was thinking of making a compiled differential evolution minimizer to complement the Nelder-Mead one, partly because NMinimize's DE implementation is very slow and implements only one DE mutation operator, and partly because NM is not suitable for all problems since it frequently gets stuck. Would anyone find this useful?

acl

10:57 PM

@OleksandrR. what are the strengths of a DE minimizer?

Oleksandr R.

@acl it is much more robust than NM. The other options are simulated annealing or random search. Random search can't be compiled since it needs to call FindMinimum extensively, and I don't consider simulated annealing particularly effective.

Rojo

So DE's strengths are that NM sucks?

acl

@OleksandrR. OK thanks. I meant more like, in which cases is it known to be efficient? I've succesfully used SA in the past (although not mma's implementation)

halirutan

@OleksandrR. Can you summarize, what DE needs from the target function?

Purely numerical evaluation, gradients, ..?

Oleksandr R.

@Rojo NM doesn't suck; it is much more computationally efficient than DE and converges much more quickly on fairly well-conditioned problems. But its search ability is also much more limited.

@halirutan black-box function, no gradients, does not need to be continuous. It just has to be defined over a continuous space.

halirutan

11:01 PM

@OleksandrR. This would be suitable for image registration and some other image related functions.

E.g. Affine image registration based on cross correlation which is usually an ugly measure.

@OleksandrR. So in essence, I could use this and write a registration purely in Mathematica.

Szabolcs

I quite like the graphics styles here: dcjones.github.com/Gadfly.jl/doc

Oleksandr R.

@halirutan I don't know anything about this problem, but yes, probably. If the objective function is fairly smooth you probably can do it more efficiently with NM, though.

A major advantage of DE is that it has better scaling to higher-dimensional problems.

halirutan

@OleksandrR. And it's a global minimizer yes?

Oleksandr R.

@halirutan yes... well, a metaheuristic. No guarantees, as usual.

halirutan

@OleksandrR. The usual stuff.

Oleksandr R.

11:05 PM

@halirutan do you know Python? I can send you a DE code in Python right now if you want.

halirutan

One last question: What about constraints?

@OleksandrR. Nope, sorry. I knew a man named Monty but that's probably not what you mean by Python ;-)

Oleksandr R.

@halirutan have to incorporate them into the objective function. Typically I just have it return a large value when the constraints are violated. DE tolerates this much better than other algorithms but there are more efficient ways of doing it.

acl

@OleksandrR. right that was what I was asking.

halirutan

@OleksandrR. Ahh, I remember you were mentioning the thing how you can easily work this into the objective function once.

Some word named after a mathematician.

Oleksandr R.

Karush–Kuhn–Tucker conditions

In mathematics, the Karush–Kuhn–Tucker (KKT) conditions (also known as the Kuhn–Tucker conditions) are first order necessary conditions for a solution in nonlinear programming to be optimal, provided that some regularity conditions are satisfied. Allowing inequality constraints, the KKT approach to nonlinear programming generalizes the method of Lagrange multipliers, which allows only equality constraints. The system of equations corresponding to the KKT conditions is usually not solved directly, except in the few special cases where a closed-form solution can be derived analytically. In ge...

halirutan

11:07 PM

@OleksandrR. OK, several mathematicians ;-)

Oleksandr R.

@acl DE is also good for functions for which it is rather difficult to find the global minimum due to many deceptive local minima.

acl

@OleksandrR. I was reading about that now. Sounds like it would have been useful when I worked for a while on a frustrated model (ie one with lots of local minima close to each other)

halirutan

11:26 PM

@OleksandrR. This is indeed a nice point.

Stack Exchange

Blog - Stack Exchange: Podcast #45 – Keeping it Sharp

Alex Miller on March 18, 2013

Our guest this week is Eric Lippert – language architect extraordinaire and famous for all his work at Microsoft in developing their languages

Eric joined Microsoft right out of college and was originally working on VB

It’s time for everyone’s favorite game: Name the Worst Feature of that Microsoft Technology!

If you’re a non-programmer and still listening, make sure to email us for your free prize

Eric now builds “static analysis” programs which actually means something real when he’s talking about it …

Transcript for