@OleksandrR. Done the tutorials? No. That's not the problem. They might not know how to do the tutorials. Or even which tutorials are relevant to their interests. This isn't a problem for people like you and me. We have sufficient background to know what tutorials are relevant.

"Everyone wants quick results with minimal independent effort". Yes. Unfortunately, with commercial software, this is what is promised. Anyone who doesn't promise it is toast.

What I like about Mathematica is that it's relatively decent at delivering the impossible.

"You want to do machine learning, but you don't even know linear algebra? Great. We have a function for that."

@Searke that's a great point. I never thought of it that way.

@blochwave I don't know what I'm more impressed about. The detection of gravitational waves, or the publication of a 16 page PRL. (6 pages are the list of authors.)

@J.M. Ah, but it's one of those situations where even if I just want the first eigenvector it still return a list like Take, so I have to use First or [[1]] anyway.

I guess for larger matrices it could save computation time though.

Well at least I figured out why they use the principal eigenvector for extracting the ranks/priorities from a comparison judgement matrix. Because if you start with a rank/priorities vector and construct the perfectly consistent judgement matrix then the principal eigenvector is exactly what you started with! So now I can be satisfied with the explanation I give my friend.

RandomInteger[{1, 10}, 5] //
 Normalize@# == Normalize@First@Eigenvectors[Outer[Divide, #, #]] &

If they included that one sentence in their paper it would be much more accessible.

Some cool image feature detection code went online in Github. Will be cool to have it integrated in MMA soon ... :) Details goes (here) [newsroom.ucla.edu/releases/…

@OleksandrR. I know, it's a long letter! But so cool :-)

@blochwave, @OleksandrR sorry to barge in on the PRL discussion but I was hoping there would be more enthusiasm here over it - did you notice the "BlueGreenYellow" colour theme in the spectrograms of figure 1 :) ?

@gpap Oh I noticed!

@blochwave the detection is a technical tour de force. I had no idea the circulating optical power is as high as 100kW! It'll take me a while to go through the references and properly understand how LIGO actually works. The radiated energy of 3 solar masses is pretty amazing as well.

@PlatoManiac Link is broken.

@blochwave the noise is actually extremely small, apart from isolated resonances that the filters were designed to remove.

@Szabolcs it has a bracket at the end. Otherwise it works

Oops, right!

@OleksandrR. @halirutan It turns out that infinities and NaNs don't work with LibraryLink either. They are not displayed as Inf` anymore, but as Infinity. They work for many purposes. E.g. unpacking the array properly converts them to the symbol Infinity. But they do not work with all functions and doing arithmetic on them may cause weird effects. For example they are Rounded to 0.

@blochwave when the resonances are removed, the instrument noise floor is between 20 and 40dB below the signal peak

@PlatoManiac Looks like it should be implementable in Mma: github.com/JalaliLabUCLA/…

@OleksandrR. Did you already read the full PRL paper?

@Szabolcs no, I skimmed it. Although easy to read, it's quite hard to understand without following up the references

@Szabolcs There is a matlab code too : mathworks.com/matlabcentral/fileexchange/…

@PlatoManiac That's the same one I linked, posted on a different site.

It will be great to see if we can bit the Matlab speed ;)

@Szabolcs note: "The software is protected under a US patent."

They don't provide any licence for the MATLAB code. They also seem to assign their copyright to The MathWorks Inc? The latter must surely be a mistake. Anyway, reimplementation should be fine I guess, as long as it carefully avoids infringing the patent

@OleksandrR. Who knows given the code is open to inspection, someone can even come up with an improved version which will not necessarily be protected by the supreme US patent...

@PlatoManiac that's not how patents work unfortunately

Yeh I know :)

Ye goats!

... with a really cool Mathematica solution.

@OleksandrR. yeah... got some spare solar masses to convert, by any chance? Kind of takes it out of your average matter budget...

@MikeHoneychurch and others?

Is any of you using MaTeX? Do you have any feedback about it? Suggestions for changes?

google.com/custom is dead. I feel like Google kills off a lot of my favorite functionality.

@Szabolcs only occasional, but very appreciated use here to touch up publication figures.

Works as advertised, no gripes here ;-)

If you care to look that up, the new Origami 6 (books.google.de/…) features one of these on the bottom of page 588.

...mind you, it is just a few characters.

@YvesKlett How excellent that the google preview only goes up to page 490 :))

@halirutan hehe... not sure how that works. Depending on your IP, login etc. one seems to get to see different pages. I put 588 in the search line, and that did the trick...

(anyway, it is just the letters A, B, C and D pretty printed with help of MaTeX)

Same here I can only see to 409 or so.

Is there any way to coerce ReadList to read a CSV file?

Is there anything I can do to tell it that , is the separator in this data?

ReadList[StringToStream["2,3\n4,5"], {Number, Number}]

I am looking at reading large CSV. Mathematica is very very bad at this compared to competitors like Python or R. I am hoping I don't have to go to LibraryLink again. I am thinking of datasets like this, a 2 GB CSV file.

Is anyone here working with Workbench currently?

@Szabolcs I did some research on this, since I need a fast general CSV reader for my work. The short answer is that it is very hard to make a correct and general CSV reader. What I was planning to do (and still am, but was busy with other things) was to use Java and pick one of Java implementations - there are several very decent.

@Szabolcs J/Link is quite efficient at data transfer, as long as data is represented as an array of primitive types, so this can work with some additional code. I plan to return to this problem rather soon, but probably not right now. Of course, you can look for C implementations for CSV reader, but writing a decent LibraryLink - based general CSV reader would probably be much more work still, than using Java.

@leonid My experience with Java based importers is that they take an enormous amount of memory and that even increasing the Java heap considerably doesn't satisfy their memory hunger.

The XLS importer is one example. My usual approach of a large XLS file is to convert it to CSV in excel which I then read with ReadList. The XLS importer more breaks more often than not and is very slow to boot. Loading times for files that just fit can be as long as 20 minutes or more.

@SjoerdC.deVries Modern versions of Java are almost speed-equivalent to C, and memory consumption should be within a factor of 2-4 worse than in C. I think you just had a bad luck, meaning that the particular converters you were dealing with were sub-optimal. Java is used in a number of commercial big-data applications, including stock trading systems, big-data analysis, etc. If there were fundamental memory-related issues with Java, that would've shown up in those.

@LeonidShifrin I was actually referring to Mathematica's importer.

@SjoerdC.deVries Yes, I understood that :)

@LeonidShifrin so, does Mathematica use a public domain importer?

@SjoerdC.deVries This might be true, but frankly I just don't know. Never checked that.

@SjoerdC.deVries To Mathematica's defense, it seems that there aren't many good XLS importers available, at least in the public domain - unlike for CSV. And implementing such an importer in-house can be pretty expensive - although might make sense given how ubiquitous it is.

@LeonidShifrin the complexity of XLS is orders of magnitude larger than that of CSV, so appreciate the complexity of writing an importer. Nevertheless, a file that effortlessly fits into Excel, should also fit in Mathematica imho but it often does not.

@SjoerdC.deVries Yes, I agree.

I think I read somewhere that the formal description of XLS runs over 1000 pages...

@SjoerdC.deVries This probably explains why there aren't many decent XLS importers around.

@LeonidShifrin yup. BTW How are you doing nowadays?

@SjoerdC.deVries Well, could be better, but better than before. Thanks.

@SjoerdC.deVries How about you?

@LeonidShifrin I'm doing fine. Kids and wife same. A bit busy at the moment, but quite manageable. Nothing compared to you. my firm is laying of a significant part of its workforce but my department will grow.

So, count your blessings, as my mum always said.

@SjoerdC.deVries Yeah, that's right. I am glad things are going well for you and your family. Hope those lay-offs won't affect you.

@LeonidShifrin @SjoerdC.deVries Hi, any of you guys using Workbench? I'm interested to see what CurrentValue[$FrontEndSession, "StyleSheetPath"] gives after Run Project as Wolfram.

Big changes going on in banking due to automation. Almost nobody needs local offices nowadays, so those are the first to go. Risk management, my branch, on the other hand has to deal with an increasing amount of regulation.

@Kuba alas, I'm an old fashioned Frontend user

@Kuba I do use Workbench, but it's been a while since I ran a project through Workbench. I run all projects via FE, and use WB only as a text editor.

@LeonidShifrin The same here, and at home I have no Workbench at all. @SjoerdC.deVries I'm interested in modyfing StylesheetPath but to no avail so far. I suppose Workbench has to do something valid so the project/FE/StyleSheet is added, and it's styles are found.

@Kuba I'm sorry, but I don't think I can help you here.

@Kuba Ok, I've run it. Here is what I get:

{"/Applications/Wolfram Workbench \
3.app/configuration/org.eclipse.osgi/bundles/14/2/.cp/\
MathematicaSourceVersioned/Head/MUnit/FrontEnd/StyleSheets",
 FrontEnd`FileName[{$RootDirectory, "Applications",
   "Wolfram Workbench 3.app", "configuration", "org.eclipse.osgi",
   "bundles", "11", "2", ".cp", "MathematicaSource",
   "DocumentationTools", "FrontEnd", "StyleSheets"}], ParentList,
 FrontEnd`FileName[{$UserBaseDirectory, "Paclets", "Repository",
   "DocuTools-0.9.1522", "FrontEnd", "StyleSheets"},

@LeonidShifrin Thanks, I'm not sure if that can tell me anything :-/

I'm trying to add desktop there so stylesheets from there are found

but this is not working

@Kuba Alas, I have no idea why.

@LeonidShifrin ok, thanks anyway :)

Sure, np :)

@SjoerdC.deVries, @Kuba Gtg now. Talk to you later!

Good night

@Kuba night.

Bye