Wolfram Mathematica

@Jin Framed@Quiet@
Blur@ImageAdd[
ColorConvert[
Erosion[ColorNegate@(MorphologicalComponents@(EdgeDetect@#) //
Colorize), 3], "HSB"],
RemoveAlphaChannel@ColorConvert[#, "HSB"]] &@
Import@"http://www.gravatar.com/avatar/\
70ad2c2830eb9a7753bd6312f3811e3e?s=128&d=identicon&r=PG"

Rojo

@belisarius is that a virus?

belisarius

@Rojo Jin colorized

Rojo

It's nice to see he got the software, hehe

Haha, jin colorized

I gotta do something about my gravatar

@magma Did you use other similar software?

magma

time to go to sleep. very late here. Good night

belisarius

@Rojo I will pay you $10 if you let me make your avatar

and you use it after

magma

12:37 AM

similar to system modeler?

Rojo

@magma Yeah

@belisarius I'll give you a 10$ discount if I can reserve my veto

magma

no

belisarius

@Rojo no chance

Rojo

@magma and what's your area?

magma

like yours. i am an Electronic engineer

with a keen interest in pure math and physics

good night

Rojo

12:40 AM

Night magma

magma

:-)

Rojo

@belisarius 10$ discount and I trust you for a day

belisarius

@Rojo The making of an avatar is an artistic moment. You are betraying art and friends by being so suspicious

Rojo

@belisarius Third world habits

A week!

belisarius

Nahh ... making the avatar you deserve will take me more than a week. A full life, I believe, suffering the stressful state of not getting to perfection

Rojo

12:47 AM

Haha

belisarius

en.gravatar.com/userimage/20688481/…

Rojo

Haha

@belisarius Better than having Peron's face

belisarius

@Rojo Bonavena's is not a bad idea

Rojo

Haha

Grr, I just hanged my session while trying to do the most basic image processing ,I suck at that

belisarius

Rojo

12:53 AM

I'll never let go of my veto

belisarius

you coward

and I still haven't let my imagination flow

@Rojo For example. You could try this for a week

just to find your real self

Rojo

Hehe

I'll put something provisional

R.M

Rojo

HAHA

belisarius

circuits overload!

I love those four eyes

@R.M Those are three iconic Argentinian actors

R.M

1:03 AM

Now it's just one super iconic... thing!

belisarius

@R.M Perhaps the woman is too iconic

google.com/…

Rojo

Now, to change the picture I have to create a gravatar account?? I thought it was just uploading or putting a link

R.M

@Rojo yes. You upload at gravatar.com

belisarius

Making your own avatar is a rite of passage

Rojo

...and now.....

...how to link the gravatar account to this?

R.M

1:14 AM

If you've used the same email in both places, it'll auto refresh

The gravatar is linked to your email

Rojo

Ok

We'll see tomorrow, I just put a plain red

If it updates, we'll see what's the definitive one :)

R.M

Rojo, Argentina

11.6k 23 52

belisarius

How creative!

R.M

indeed

belisarius

You could Inset a Big "R" and win a Guggenheim grant

Rojo

1:18 AM

Back, now it's updated

I did it in Paint

rcollyer

I'm looking to retag this question. Obviously, it shouldn't have vector as it is discussing std::vector not $\vec{x}$. But, maybe c++? Maybe not.

belisarius

@rcollyer I think C++, perhaps mathlink

R.M

c++ might be good... We have fortran, python, java, C#, etc.

rcollyer

That answers that. But, list-manipulation? I don't think so. Nor, osx.

belisarius

@rcollyer agree

rcollyer

1:22 AM

Any objections to my removing those to also?

belisarius

nope

rcollyer

done.

R.M

interoperability, c++, output-formatting, mathlink should be sufficient, I think.

rcollyer

why the last one?

it isn't output, per se.

R.M

I went by this paragraph:

> I want to store each of these vectors as a list in Mathematica, so, for instance, if three iterations returns the vectors (0, 1), (0.02, 0.9) and (0.04, 0.73), I want this output to be written into a Mathematica list of the form {{0,1},{0.02, 0.9},{0.04, 0.73}}.

I guess output-formatting is more for NumberForm, etc.

rcollyer

1:25 AM

okay. 6 of one, half a dozen of the other by my thinking.

But, adding interoperability.

Oleksandr R.

@Rojo now you look like the Soviet Union, or China. But smaller, obviously.

Rojo

@OleksandrR. Yeah, I'm not liking it much. Too aggressive

rcollyer

@OleksandrR. careful, he may get an inferiority complex, and before you know it he'll be threatening Japan and the US. :P

Oleksandr R.

@belisarius I would say more like a right nuisance! I'm not completely happy with my current one but it clashes with the site design less than the previous one did, so I can't be bothered to change it again.

belisarius

@OleksandrR. :)

Oleksandr R.

1:30 AM

@Rojo as I understand it, he's referring to the fact that in many cases, attempts to change a part of a data structure will result in the entire structure being copied and the part changed only in the copy, leaving the original unchanged. ReplacePart does this, for example.

belisarius

Anybody knows how MaoTsé entered my monitor?

Rojo

@OleksandrR. Thanks. What I'm actually interested in understanding well is, what does he(/you, etc) think would be nice to have. I've read so much about it that I want to give it a shot

and I trust his judgement on what's missing and would be useful

For example,

Having a function to "subscribe" a symbol as a copy of another, and a custom set that keeps both up to date when set

without having 2 copies of the data

and having the data remain on the other symbols if you clear one or set one with the regular Set

and no leaks

That would be enough?

Oleksandr R.

Well, that sounds like it could be useful, but IMO the problem in general is more one of efficiency than semantics.

Rojo

The thing is, I think that "that" could be done quite simply

Haven't tried it yet, but

if it worked, it would as inefficiet as MMA is in general

Oleksandr R.

Yes. I understand where you're coming from, but it still doesn't buy you in-place updates. Built-in functions still copy extensively.

Rojo

1:39 AM

Definately

Built-ins grr

So, in conclusion

Oleksandr R.

For example, someone asked me a while ago if it was possible to do an in-place FFT. I didn't follow up on that but I am certain it is not without some serious messing with LibraryLink.

Rojo

What he means is something that can't be helped as long as one can't make built-ins work in -place

Oleksandr R.

I believe so, yes. Of course one can use things like linked lists which have much less to copy and thus have fewer performance pitfalls.

Rojo

Yes, built-ins copy a lot, but

if you have a long list, say, "l"

and you do l+1

Oleksandr R.

There is some object called a "raw array" which I believe is basically used when calling Fortran (LAPACK for example) because of the Fortran semantics. However how it works I don't know.

Rojo

1:44 AM

You are right

You have RawArray, Dev`FromRawArray, RawArrayQ. Had no idea

So, list+1

or, worse

++list

while it is evaluating, it uses up memory as if copying the list

but, does that imply it's slower that it would be in place?

Yes, right?

Because it doesn't mean it copies the list... Probably reserves the space and puts the results there

instead of in the original

I'm just trying to finally understand this big issue a little better

Oleksandr R.

Well, that still involves a copy. It could update in place. Although, there aren't really semantics to express that.

Even the compiler copies extensively, it would seem.

Rojo

I see that there's copying in functions that modify only a little part of a possibly long expression

such as ReplacePart

(which is one of the few that can be reimplemented in place)

(with part)

Oleksandr R.

I think fundamentally it's hard to mix functional/term-rewriting semantics with mutable data structures without making a mess of things.

Part is one of very few exceptions...

I think even Part may copy in some circumstances though. For instance, when operating on a packed array.

Rojo

So far, the limitation I really see is the one of not being able to extract a subpart of an expression (say half of a long list) without making a full copy of it

Oleksandr R.

Yes, that's a major one. Of course you can use linked lists in that case.

Bags are also possibly useful in that they can be passed around like pointers to their contents.

Rojo

1:53 AM

Any different from passing a Hold?

Oleksandr R.

Well, you can destructure Bags without (necessarily) copying. Of course if you call BagPart[b, All] then it copies.

As I understand it a Bag is simply a very low level linked list.

Rojo

... a linked list with heads that hold, that allows you to extract parts of it easily (not like a linked list), but doesn't let you remove elements

and that's actually an atom

which who knows why should I care

Oleksandr R.

I'd consider that semantically the same as a Bag.

Rojo

I was describing a bag

Hehe

Oleksandr R.

Oh, okay. :)

Rojo

1:59 AM

Thanks for the Heap ;)

Oleksandr R.

Did you make any use of it? I haven't tried, yet.

Rojo

Not real use. Fake use, when I tested it a little bit

Oleksandr R.

Heavens to Betsy!

0

Q: Integrate and NIntegrate problem

exp := -5.082019443309081E^(-0.7036967536123101 (2/25 - (2 x)/ 25)) (E^(-0.11728279226871835(32/25 - (32 x)/25) - 0.11728279226871835 (2/25 - (2 x)/25)) - E^(-0.2345655845374367(17/25 + (8 x)/25)) (1 + 0.2345655845374367 x))^2 x (5.515552687747721\ E^(-0.23456558453743...

numerical-integration

Rojo

I'mi gonna upvote that

belisarius

2:12 AM

It has been a long time since I saw a question that I am so glad to pass on

Rojo

"I have a very long integrand,"

3

Hahahaha

belisarius

@Rojo Are you boasting?

2

Oleksandr R.

@Rojo Hmm, I didn't mean it that way. But maybe this poster deserves it. :P

Sjoerd C. de Vries

@verbeia shouldn't this be a comment?

0

A: Integrate and NIntegrate problem

It's hard to work out the problem when the code creates syntax errors once pasted back into Mathematica. However, I would point out that you have some very long floating-point numbers multiplying exact numbers, all taken to the exponent. Integrate will use numerical methods anyway in that case. ...

@belisarius Rotfl

belisarius

:)

That formula is the nearest thing to spam maths could possibly be

R.M

2:21 AM

2

must be a standard warning on any question longer than X

belisarius

starred!

rcollyer

@belisarius would you please remove your comment on his question. Thanks.

Sjoerd C. de Vries

@rojo My integrand is bigger than yours

belisarius

@rcollyer Oh well, but I warn you Mma 9 has amazing new msgs

rcollyer

@SjoerdC.deVries Simplify will take care of that problem. :)

R.M

2:26 AM

@SjoerdC.deVries Too bad, it doesn't grow with x

Sjoerd C. de Vries

Hehehe

rcollyer

@SjoerdC.deVries Depending on the method used, Integrate[exp ...], Integrated[Expand ...], Integrate[#, ...]& /@ Expand[exp], etc. I get different answers. So, I think she has the correct answer.

Rojo

@SjoerdC.deVries I pity you then

It is actually interesting to ask you all

rcollyer

How big our integrands are?

Rojo

No, not really, we have ladies in the forum

Sjoerd C. de Vries

2:31 AM

;-D ;-D

Rojo

How to know what's the form of an expression that leads to less numerical errors

rcollyer

That is an interesting question, and I wish I knew.

Rojo

I'm not talking about integration any more

Sjoerd C. de Vries

Horner form?

R.M

@Rojo don't work with exp(big number)... instead try to work with log and then take exp

belisarius

2:32 AM

Is there an easy way for NDSolve to refine its range (ie make it smaller) based on a condition on the derivatives?

NDSolve[
{y''[t] == ... && y'[t] > 0 ...}, y, {t, 0, 1}]

Rojo

@R.M Why is that?

R.M

There was that question from that physics.se optics guy

Sjoerd C. de Vries

There are two close votes on the question. What do you all think?

It's a bit specific

belisarius

@SjoerdC.deVries three

rcollyer

3 votes, now.

mine.

belisarius

2:36 AM

mine

and phantomas

rcollyer

@belisarius no, it's mine! :)

@SjoerdC.deVries I think it is too localized.

Oleksandr R.

Voted NaRQ. The integrand is a mess and Integrate's assessment of it is totally dominated by numerical error at machine precision.

belisarius

and Mma::psycho, who was forced to left the party

rcollyer

yeah, but he's just no very constructive, or conducive to learning.

Hey, Sjoerd didn't even have to touch it.

R.M

@Rojo The exponentials blow up: Exp[10.^16] Exp[-10.^16] vs Exp[Log[1]]==1

rcollyer

2:40 AM

(For some strange reason, my mind keeps going into the gutter.)

@R.M Also, you're effectively using the series form of Exp which increases the number of multiplies dramatically. Even in Horner form.

Sjoerd C. de Vries

@rcollyer The art of delegation...

Rojo

In an unrelated issue, 100MBs is approximately the max bigint it's saving witout complainig

R.M

30

Q: Is it possible to make Mathematica reformulate an expression in a more numerically stable way?

I'm writing a numerical optimization, and I'm having a problem with an expression of the form $$ e^{-t} (1+\mathrm{erf}(t)) $$ The overall shape of the function looks correct, but when $t$ is small, $e^{-t}$ is huge while $(1+\mathrm{erf}(t))$ is very small, and their product is also small. This...

numerics simplifying-expressions

@Rojo ^

Rojo

@R.M Niiice

rcollyer

@SjoerdC.deVries I figured. We're piranha and all.

And, someone chummed the waters!

belisarius

2:43 AM

@rcollyer We must bully someone from time to time. If not, we risk eating each other

rcollyer

@belisarius as I said:

Rojo

@belisarius Let's visit a Maple forum

rcollyer

4 mins ago, by rcollyer

(For some strange reason, my mind keeps going into the gutter.)

Sjoerd C. de Vries

Anyone planning to go to the wolfram tech conference?

rcollyer

@Rojo or matlab.

belisarius

2:44 AM

@Rojo Hey! That's a NICE idea

rcollyer

@SjoerdC.deVries I've been thinking about it, but I need to get funds for it.

Rojo

@rcollyer R.M is about to write a blog post to evangelize Matlab users, and belisarius hates Maple

Sjoerd C. de Vries

@rcollyer SE can do that

R.M

1

Maplesoft

Proposed Q&A site for users of all experience ranges with Maplesoft's numerical analysis software.

Currently in definition.

belisarius

@SjoerdC.deVries Can rep points be cashed ?

R.M

2:45 AM

Join!

Oleksandr R.

@Rojo are there any? I wondered that the other day but was unable to find anything like what we have for Mathematica.

rcollyer

@SjoerdC.deVries that's true. Maybe a moderator should bring that up with the powers that be? :P

Rojo

;D

Let's joiiin

R.M

@rcollyer Jin said that they are fully aware of the conference and they do plan to send some members of the community

Sjoerd C. de Vries

16

Q: How can I get Stack Exchange to send me to a conference?

I read Supporting Community Conferences, which states, Depending on the circumstances and location, we can also sponsor community leaders to attend an event on behalf of their site. We will subsidize your costs to attend, within reason, and provide you with a bunch of swag to use as an ice-br...

discussion faq site-promotion speakers-bureau

Rojo

2:46 AM

1 follower, sad

rcollyer

@Rojo as long as you're trying to "save" the matlab users, then sure.

belisarius

@R.M You are a professional agitator

R.M

Jul 13 at 3:10, by Jin

I'll have our community team make a meta post about this, to see who's all interested in going

Sjoerd C. de Vries

Aarthi indicated that SE would indeed be willing to sponsor this

Rojo

@rcollyer You used Matlab?

Sjoerd C. de Vries

2:49 AM

I attended last year, fully self-funded and had to take vacation

rcollyer

@Rojo I was in a class where the professor demanded that we use it despite the fact that I copied the functionality we were supposed to use in mma in under a 1/2 hour.

Before I nominate myself to go, I'd have to check a few things.

belisarius

I would like to attend. If you @All support my candidature I promise bringing back an autographed copy of NKS v2 for each one

rcollyer

(I really have to get my mind out of the gutter!)

Sjoerd C. de Vries

Bring back a copy of v9 for us and I'll support you

R.M

You can either go to the conference or get v9... hmm. Tough choice

Rojo

2:51 AM

I'm happy enough with the MMA9 documentation

rcollyer

All of us, every user with 500+ rep

Sjoerd C. de Vries

I suppose v9 will be announced there

Rojo

Let's hope

rcollyer

likely. depends on if they get the testing done and the features in in time.

belisarius

@SjoerdC.deVries mmmm ... ok. I'll give you Maple v9 copy

R.M

2:52 AM

@rcollyer Unlikely that they'd be working on features now, no?

rcollyer

@belisarius aren't they on version 15 or 16?

belisarius

@rcollyer No idea really

R.M

@rcollyer Maple9 is the latest in Argentina

Oleksandr R.

Maple is at 16 currently.

Sjoerd C. de Vries

That's matlab I'd say

Rojo

2:52 AM

Uh, and we are at 8. Half

R.M

Meh. (old) Firefox vs Chrome, really.

Sjoerd C. de Vries

I forgot, why is my name in blue here?

R.M

mod

rcollyer

@R.M depends on a lot. If they're releasing in Oct., then putting in new stuff now would put them behind. But, they may just be announcing when it will be ready for sale in Oct.

Sjoerd C. de Vries

I really have to get my mind out of the gutter

R.M

2:54 AM

@rcollyer That sounds more like Apple's strategy...

"World's most advanced OS, blappity blah Mountain Lion... coming this July" — announced in June

rcollyer

@R.M I have no insight into their strategy, so I'm just guessing.

Sjoerd C. de Vries

I've heard rumors beta testing is underway

In beta features are frozen, right?

R.M

yup

rcollyer

@SjoerdC.deVries They should be. But, not if they go the netscape route.

belisarius

We could blackmail the SE team to force them to send all of us to the conference. If not, we will start posting the "long integrand" question on all sites

rcollyer

2:57 AM

But, being hellbanned from the entirety of SE wouldn't be worth it. We'd get more stuff done, though.

R.M

alright, gotta head home now.

Adios

rcollyer

See ya.

belisarius

@R.M Auguri

Sjoerd C. de Vries

@belisarius The meta post says "community leaders", which sounds sufficiently plural to me.

rcollyer

@SjoerdC.deVries true, but you have to define "leader."

belisarius

2:58 AM

@SjoerdC.deVries communist leaders? Yes

Sjoerd C. de Vries

@rcollyer Bye

@R.M Ciao

rcollyer

@SjoerdC.deVries I was saying bye to RM.

belisarius

@rcollyer bye!

ohh my

rcollyer

@belisarius copy cat.

belisarius

purrr

Sjoerd C. de Vries

2:59 AM

Harhar

Thunderstorms here

rcollyer

@belisarius Well, that hits my limit of not being able to respond due to it being a public forum!

So, good night all. I have to go beat on a cluster.

belisarius

@rcollyer You shouldn't use your real name

Rojo

Bye

Sjoerd C. de Vries

Bye again

belisarius

I am not saying goodbye again

:P

Sjoerd C. de Vries

3:01 AM

@belisarius So, your name isn't real?

belisarius

@SjoerdC.deVries Ohh yes

This is me

Belisarius

Flavius Belisarius (, ca. 500 – 565) was a general of the Byzantine Empire. He was instrumental to Emperor Justinian's ambitious project of reconquering much of the Mediterranean territory of the former Western Roman Empire, which had been lost less than a century previously. One of the defining features of Belisarius' career was his success despite a lack of support from Justinian. He is also among a select group of men considered by historians to be the "Last of the Romans". Early life and career Belisarius was probably born in Germane or Germania, a city that once stood on the site ...

only younger

Sjoerd C. de Vries

What's in a name

That which we call a rose
By any other name would smell as sweet.

belisarius

Umberto Eco would be proud of you

Sjoerd C. de Vries

He is

belisarius

Tell us more

Oleksandr R.

3:06 AM

@SjoerdC.deVries Not if you called 'em stench blossoms.

Sjoerd C. de Vries

I managed to finish Foucault's pendulum...

belisarius

@SjoerdC.deVries I enjoyed the name of the rose a lot more

Sjoerd C. de Vries

@belisarius The pendulum started ok, but the plot totally derailed in the final chapters.

belisarius

@SjoerdC.deVries I've read "The name.." in two days or so. The "pendulum" took me a month

Sjoerd C. de Vries

I forgot whether I read it or saw the movie

belisarius

3:20 AM

@SjoerdC.deVries Oh ..the movie wasn't up to the book

Verbeia

@SjoerdC.deVries I thought it was too long. Can I even remove the answer once the question is closed? I posted it before others started commenting about numerical issues.

Sjoerd C. de Vries

@Verbeia I was just wondering. Rcollyer thought it was Ok, and the point is moot now anyway.

Verbeia

@SjoerdC.deVries If the community feels strongly, I'm happy to remove it.

Sjoerd C. de Vries

@verbeia that's not the case. So, no problem. You might want to scroll a few pages back to see the discussion that went on.

Getting some sleep now. Bye!

Verbeia

3:39 AM

@SjoerdC.deVries I just did, and I agree, this sort of code dumping should be discouraged.

But I think the way to encourage good behaviour is to tell them what they are doing wrong, and still give them some help, if not complete. If they are repeat offenders, by all means come down harder. But this user has another question with 10 votes, so I think they are amenable to polite education.

Transcript for