« first day (1309 days earlier)      last day (3172 days later) » 

3:16 AM
@SjoerdC.deVries Huh?
 
3:52 AM
@kirma, read the first sentence again. Does it remind you of anything? :)
 
Heh. :)
 
 
7 hours later…
10:41 AM
@SjoerdC.deVries don't be so mean... SW has a marketing duality problem. He drives what is probsably an almost 100% commercial business, while having many of his clients in the academic world. Every day we see adds from cars and TV manufacturers that have a selling language WELL above the "new kind" of level (some have invented the colours!, others, you spray something on you and you become a human magnet!, others are "probably..." the best in the whole UNIVERSE!, etc).
And these techniques are used by companies of technical software similar to Mathematica (just look at what CFD companies are saying; I'm almost sure that some of them have invented the flowing notion itself!, or participated on the revolution of your car performance with a single simulation!). And rest assure that this discurse is also coming directly from the mouth of the CEO, and not not just printed in fantastic colours. And these softwares are also extremely used in the academic world!
But since they are closer to the applied disciplines, and not used mainly by their more pure academic counterparts (don't take it the wrong way, it is just a different approach to production and a different set of driving forces), they are "saved", while "our" SW is constantly under attack...
(may this compensate the ethics all my complains on functionality and bugs frustration :-)
(Sjoerd: no personal attack intended. I just profited from the occasion to write some general stuff on this subject...)
 
11:01 AM
@P.Fonseca it was just a reference to the title to his book "A new kind of science". He has been using this "new kind of" phrase on various occasions.
 
11:25 AM
@SjoerdC.deVries I understood...! It is his registered "next big thing" phrase.
 
 
2 hours later…
1:27 PM
Unfortunately Google Trends doesn't go back to NKS publication to see if it had much effect. Davic Cameron referred to a "new kind of country" though.
Book mentions probably aren't a good metric for marketing terms, but it has declined it seems.
 
@MichaelHale It's just when it's being said a lot by the same entity that it's noteworthy. Certainly it's not a new kind of expression.
 
@Pickett Right, but if it was a successful marketing tactic, you would expect to see it trend up as others adopt it.
That being said, I've spent a lot of time reading his book just because I keep finding it interesting. Went on a binge yesterday for the first time in a few months.
 
1:47 PM
@MichaelHale Well, Fonseca didn't say that that expression is uniquely good. The thing is that in PR speak revolutions are everyday occurrences.
 
@Pickett I'm not criticizing Fonseca. I was just curious and posted what I looked up.
 
@MichaelHale I want to read that book, someday. Probably won't get to it, though.
@MichaelHale It would be more interesting to know if any particular expression see clear trends like that. I don't think that's how language works... ads try to be different, not all the same.
 
@Pickett Maybe not, but I think they copy successful tactics all the time. Hence charging $1.99 instead of $2 is so popular.
 
@MichaelHale Absolutely, just not the exact same expression. They'll try to write it so that it sounds even better.
 
@Pickett I think if the book became a runaway success of lasting cultural significance that there would be definite trends related to the phrase. Think of how many times you've seen "A tale of two __", "To __ or not to __", etc. I was just presenting those charts as evidence that it hasn't. I guess that's obvious though because I'm the only person I know that reads it, and the only time I see people discuss it is in this chat room or in an official Wolfram forum/event.
Ah. They are called "snowclones".
 
2:10 PM
+1 for snowclones, spot on.
 
 
2 hours later…
4:45 PM
Suppose we have something like a Fermi function (Plot[1/(Exp[x] + 1), {x, -10, 10}]) in multiple dimensions. That is, the function goes from one saturation value to another according to such a function when crossing some contour in N dimensions. Other than the property that is saturates at two distinct values, the function can be quite general (the shape of the contour, the speed of saturation can vary in different ways).
What is a good numerical method to search for the roots of such a function? @Guesswhoitis. ?
Methods that rely on the derivative (such as Newton's method) don't work well because the derivative is near-zero on the plateaus. Bisection doesn't work in multiple dimensions.
 
@Szabolcs "The roots" ?
 
I'll give a concrete example.
 
ok, brb
 
f[x_, y_] := 1/(Exp[(x^2 + y^2)/5 - 8] + 1)
g[x_, y_] := 1/(Exp[3 (x - 3)] + 1)
Find x and y so that f[x,y]==1/2 and g[x,y]==1/2.
 
ah, ok
 
4:55 PM
Assume that f and g cannot be computed to high precision thus the derivative at e.g. {x,y} == {0,0} is useless.
 
ok, I understand. Nice problem
@Szabolcs How many dimensions, BTW?
 
In practice, 2-6 dimensions. f and g may also be fairly expensive to compute so minimizing the number of evaluations is important. Blindly computing them over a grid to find a point where we're not on a plateau is probably too expensive.
 
@Szabolcs hehe, that's always the first thing that comes to mind
 
Here are plots to make it easier to understand what my problem is:
@xslittlegrass It should always use the same resolution that's used for display on-screen. If you have a retina Mac, it's going to use non-retina resolution.
 
@Szabolcs In the "temperature"->0 limit (ie a discontinuous function) you can't do anything better than a clever bisection. So you'll need to exploit some property of your functions to do better than that
 
5:04 PM
@belisarius Then perhaps the question is: how to best generalize bisection to multiple dimensions.
 
@Szabolcs May be. Also interesting.
done :D
Who lost more rep due to rep cap in this site?
Any way to calculate that using Mma?
Perhaps I should post it as a question. In main or in meta? Opinions?
 
5:58 PM
posted on August 19, 2015 by Todd Rowland

For three weeks this past July, Wolfram held the annual Wolfram Summer School for over 60 students from around the world. They came to work on projects ranging from aperiodic hexagonal tessellations to computer language grammars to political sentiment microsites. The overarching theme was entrepreneurial science. Participants employed cutting-edge computational tools like Wolfram Programming Cl

 
@belisarius I'm pretty sure it's Mr. Wizard, because he's the only one who has the legendary badge.
 
@Pickett Not necessarily. The buttocks answer may have lost quite a lot in two days
as an example, I mean
 
@belisarius That's why I said "I'm pretty sure" instead of "this is how it is". But Mr.Wizard got that badge in 2013, a no one else has that badge yet. So I expect his lead to be big enough, to be above the kind of events that you mentioned.
Still, of course it would be much better if it was possible to calculate it!
 
6:14 PM
@Pickett Agree. But IDK how :) Perhaps the stackexchange data repository is enough
 
6:46 PM
@belisarius I wrote this code, I'm not sure that it gives me the exact right number, but it tells me that Mr.Wizard lost 5308 points due to being over the limit, and that seems reasonable. However I will run out of quota if I run it on all members.
 
@Szabolcs, best derivative-free multivariate method that I know is Brent's principal axis method. Mathematica has an implementation of this.
 
@Pickett Nice!
@Pickett I'm still on V9, so I can't run it :(
 
@Szabolcs, if desperate, you can form the sum of squares of the equations and use something like Nelder-Mead to minimize the resulting objective function.
 
7:02 PM
@belisarius I ran it on your ID and got 2561. For kglr I got 940. Halirutan has that projection of image onto handlebar question in his history, which is a good example of what you mentioned earlier and for him the function returns 4072.
One day halirutan lost 630 reputation points for example, probably because of that question.
 
@Pickett Awesome
 
7:22 PM
I'm 10rep off my first 200rep limit, which is a bit excruciating :P
frantically waiting for new questions to come up so i can snipe them
hooray! thanks to whichever wonderful person topped me up
 
@PatrickStevens There you go
 
@belisarius a shiny new badge for me - thanks!
 
@PatrickStevens It wasn't me. I was at it, but someone was quicker
faster?
 
@belisarius either works
 
@PatrickStevens :D
 
7:33 PM
It was me, trying to create a market for my newly written script :)
I only have three such days myself :/
 
@Pickett Ha! The free self regulating market attacks again
 
:)
 
@Pickett Excellent - when this has happened twenty more times, I'll be sure to make use of it
 
7:48 PM
Great!
Everybody, I have developed a website that makes it really easy to share Mathematica packages. I'm about to start refactoring... and still thinking about if anyone actually wants to use something like this. Any ideas are welcome, the screenshot below explains how it works.
3
It will also be possible to implement a package manager in Mathematica, like what exists for other languages.
Contingent on it being finished it will be released as open source, and I will also host it on some yet to be determined URL.
 
8:16 PM
@Pickett Looks great!
Is it possible to simply add links and maybe a short description instead of upload a package?
Many package authors would prefer hosting at a single location instead of multiple ones. Also, if links are allowed, then others than the package author can post packages.
A so far theoretical question: if someone other than me posts a package I made, do I have the possibility to edit that post myself?
 
8:56 PM
@Szabolcs Those are hard questions to answer, to be sure. None of that is set in stone yet. What do you feel is the right approach?
 
@Pickett it would be nice if your site were Git Friendly. So I can put my package into git, but register it in you site. It would be nice do have something equivalent to cran.r-project.org in Mathematica.
* your site
 
@Murta Yes, that's the type of website that I want to create for Mathematica. rubygems.org and npmjs.com are two sources of inspiration. pkg.julialang.org is an example of a system that consists only of links, and all packages are on Github. This works for Julia.
Now the question is what version or mix of these types of antecedents will work best for Mathematica.
 
@Pickett I lost 100 rep just today. Life is unjust.
 
@belisarius Ouch!
How?
 
9:11 PM
@Pickett By rep cap
 
Right, of course. At least that's a problem I don't have :)
 
@Pickett :D
 
But ok, so there will need to be a way to "link" to packages (on Github, say) instead of uploading them. However I still want to require people to upload version information etc. so that it's possible to build a package manager.
 
9:28 PM
@Pickett I was toying with the idea to create such site for a few years now. My final conclusion was that it will only really work for Mathematica if there is an easy-to-use UI (Mathematica client), that would integrate a few components, such as package manager, primitive VCS, and an ability to both publish and load packages in "one click" from Mathematica.
@Pickett I have written an (almost complete) VCS based on publishing code on Github as gists. This can solve most issues such as versioning, simple collaboration, and also automated format in which packages are submitted.
@Pickett The only missing piece was the UI, and I've been developing a FE-based IDE for the last 2-3 years. I work in it daily myself, but unfortunately the development has been very slow in the last year. Still, I was planning to move this forward to a beta stage soon. That IDE would have plugin architecture, and many things would then naturally become plugins
@Pickett That would include the package manager, VCS, and the rest of that infrastructure - for example, can also include a client to the site you have started. I really do believe in the integrated FE-based approach, because that's how the majority of users work with Mathematica. A round-trip to some external IDE wouldn't be desirable for many.
@Pickett Otherwise, I also looked at npm, and also CRAN. But as I said, I think that the thing that would unify it all and make it possible to create a smooth workflow is a good FrontEnd-based UI plus packages that would make it easy to publish and download versioned packages to the web (my VCS is based on publishing to Github gists via gist API).
 
hi
 
9:51 PM
@LeonidShifrin I remember reading about your project, but I didn’t know that it was still being developed. I started working on this after Szalbocs pointed out that there are many packages here on MMA.SE and elsewhere that are difficult to find, so for me the priority is to create an online archive first, and hopefully get that launched very shortly.
@LeonidShifrin I took the .json meta file concept from npm to be able to handle basic package management tasks from inside Mathematica, like download packages and put them in the appropriate directory and update them if there are new versions. But for me that’s just a matter of putting files in the right directory.
I'm not at all trying to change the way people develop packages, I'm just trying to provide a dead simple way of sharing and finding packages.
 
@Pickett I am not questioning your approach, the site will surely be useful. My point was rather that Mathematica has certain peculiarities, which might make the more "standard" approaches to code-sharing less effective for us. I would mention small size of the dev.community and most of its members not being professional developers.
@Pickett Having a central repo site is certainly a necessary and welcome step, but the real problem as I see it is to gain some critical mass and activity to make that collection really useful and extensive. In that respect, having seamless integration with Mathematica and simple visual way to publish, that would be integrated with the workflow, seems the key point to me.
 
10:14 PM
I have the following Matlab code:

k=1:1:10;
a=[3,5,2];
for j=1:length(k)
for h =1:1:length(a)
x=a(h);
y1(h)=x^j-2;
end
minfind(j)=min(y1);
end
plot(k, minfind)

I would like to write the same code using Mathematica. So far, I write the following:

a = ( {
{3, 5, 2}
} );
For[j = 1, j <= 10, j = j + 1,
For[h = 1, h <= 3, h = h + 1,
x = Part[a, h];
ytest = x^j - 2;
y1 = Append[y1, ytest];
]
ymin = Min[y1];
minfind = Append[minfind, ymin];
]

But it is not work.
Can anyone help me in this?
 
@LeonidShifrin I didn’t think you were. I see the same problems that you do, I just don’t have the same insight as to how to fix them - so it will be really interesting to see your solution when the beta comes out.
One thing I might say about the Mathematica community is this: it feels like there isn’t an audience to share Mathematica packages with, which may cause people to not publish them, esp. non-developers who don’t use Github. Creating the perception that there is an audience for Mathematica packages, might breathe some life into the community.
 
@Leonid @Pickett Personally, the biggest problem I would like solved is making it easy to discover packages (either for usage or for contributing). Thus the first goal is the archive. Automatic installation and package managers are the next logical step, but I fear that trying to include this from the start makes the project too complex to succeed. Planning for it from the start is a difficult thing.
 
@Pickett I actually do think that there is an audience. I just think that that audience is less prepared to use the kinds of tools and workflows that are popular among other communities - such as, for example, external code editors, command line, etc. Most people are used to work comfortably with the FrontEnd.
 
Making such an archive that can actually prevail is a difficult balancing act between simplicity and usefulness. It's too late now but I'll start a meta discussion tomorrow (if someone else doesn't start it til then :) where I'll write down what I think and why the Wolfram Library Archive is not good even as a simple archive.
Good night!
 
@Pickett Even the necessity to create a package is a pain for many of those who would otherwise like to share their code. Besides, the granularity of Mathematica code is frequently quite smaller than what really would require a package. In general, I think it would be a mistake to separate packages, their sharing, installing, searching etc from the other parts of the workflow, and here is where I disagree with @Szabolcs.
@Pickett For advanced users and experts, finding a package might indeed be one of the key problems, but in general, we just don't have that many modern and up-to-date packages for that to be a real problem, I think. The real problem for me is that I can't easily share code a a part of my routine workflow in Mathematica, and also can't download and use other pieces of code easily.
@Pickett BUT, I view the problem of searching and loading the packages still secondary, while the problem of easy construction and publishing (and versioning) of packages as a primary one. I do know for a fact that I have code for literally hundreds small packages which could be quite useful, but just because it takes time to package them and publish, I stop and not do it.
@Pickett I am sure the same is true for many other users. I also think that whatever old packages we find useful to add to the new repo, they should be tested and converted to a new format anyway, and so the creation stage is critical for them too.
@Pickett I also do believe that specifically for Mathematica, the solution has to be integrated to be really effective, because all such repositories are only effective if they have an active community around them - and for that one would have to lower the barriers as much as possible, in particular because things like command-line (which are used with the tools like npm or pip in Python etc) are rather alien in the Mathematica world.
 
10:39 PM
@LeonidShifrin I totally agree with your integration necessity. I would love to know the number of users that work on the workbench... Most of the users I know have tried it, but preferred to limit themselves to the M standard FE, simply because it makes more sense for the typical prototyping workflow that characterizes the work on Mathematica.
 
@P.Fonseca In fact, it is this integration issue and the complete lack of time in the last year that delayed me for so long - a number of tools for code-sharing I have ready or almost ready for a long time. I do hope though that the mentioned IDE I can bring to a beta-stage relatively soon, and it then should serve as a catalyst for everything else in the toolset I have in mind.
@P.Fonseca @Pickett But if the site starts to work with even the simplest functionality soon, that would definitely be great regardless of what I just said.
 
@LeonidShifrin nom and pip are a hassle for inexperienced Python and node.js users because it’s a different paradigm for them. In Mathematica this functionality would be provided through a Mathematica package, and would be used in the frontend, which is exactly what Mathematica users already know how to do.

I think the bar is low enough with a repo site. In fact I choose to provide hosting for packages because I didn’t want to require people to use Github like Julia does for their system, because that would probably be new to lots of Mathematica users.
 
@LeonidShifrin as you said, most of the users are not developers in the pure sense of the word. Most of us are developing for direct application in other type of production work. So it does make sense to add the tools to the already existing workflow, instead of to convince the user to use other tools. I can't wait to see the IDE functionalities you are adding! and yes, as long as there are a couple of useful packages listed on the launch, the thing should catch...
 
@Pickett I chose the middle-ground and decided to base code-sharing on Github gists, which have versioning. I have built a simple VCS on top of gist API, which allows the user to do simple version control. With the hosting, the possible problem is that the users can mess up the versions. Also, that makes everything to rely on your hosting, which means that if anything happens to it, all is gone. With Github, gists allow forking and collaboration, and there are more guarantees.
 
@LeonidShifrin Yes, I agree with all of that.
 
10:55 PM
@Pickett Re: Mathematica package - this will solve the problem, to an extent. But I still have a feeling that this would not be enough for a really effective work. I'd think that such operations as package manager commands should be somehow separated from the "usual" work in a conveneient way. In usual IDEs, I do have some console window. In my forthcoming IDE, I have tabbed windows, including console-like ones, and also package / directory explorer. But the pure FE doesn't have it.
@Pickett @P.Fonseca In general, I think that the workflow of a typical user in Mathematica is very different from that of most other languages, because it is centered around a dynamic interactive FrontEnd, not code editor. This has implications which must be taken into account when tooling is constructed around these workflows.
@Pickett @P.Fonseca All right, folks, I am off for today, will go get some sleep :). Talk to you soon! @Pickett, I hope you didn't take my words as any kind of discouragement, because they really aren't. I am quite excited by your initiative.
 
@LeonidShifrin I keep hearing about you tabbed version, and other buddies! Sounds great! Can you post a small print screen? Or is this still to draft for public?
 
@P.Fonseca Right now I am on a very slow connection, so I can't do that just technically, but let me search for a link...
 
See you all tomorrow (here's 1 am...)
 
@P.Fonseca All right, see you then. Remind me some time soon, I will post the screenshots.
Hi @Murta! Too bad I am just about to leave the room. Hope you are doing great!
 
It's 1 am here as well and I need to leave as well. @LeonidShifrin it's great to hear that you are still working on this. I look forward to more news.
 
11:09 PM
@Pickett @P.Fonseca It's 2 a.m. at my place, so yes, time to go. Re - working: well, very slowly, the last year was pretty tough for me. I do have hopes to get some more time for it soon, though. Will keep you guys in the loop.
 

« first day (1309 days earlier)      last day (3172 days later) »