[ art / civ / cult / cyb / diy / drg / feels / layer / lit / λ / q / r / sci / sec / tech / w / zzz ] archive provided by lainchan.jp

lainchan archive - /sci/ - 545



File: 1482102333195.png (152.52 KB, 213x300, 3e209529811899a18480e9d395e08cc581dc7360.jpg)

No.545

Can we have a thread for discussing how to do research? Where and how to find papers, how to read them, how to organize what you have learnt, how to keep up with changes, etc.

  No.546

Pretty sure this is the bulk of a PhD. I have my own ad-hoc way of researching but I don't think it's particularly disciplined or fruitful.

  No.547

This is basically just writing a report, just find a free use database or your preferred search engine and look up peer reviewed articles.

  No.548

>>546
by the time you are studying at that level you should know how to do these things, finding source material, digesting it, and organising it (in a pile on your desk i imagine) with the exception of keeping in the loop, that being one of the things you do as a phd.

the real strength of the usual academic route is your injection into the research community, your supervisor(s), colleagues, other researchers, seminars, etc.

  No.549

>>545
I think such in-depth advanced topics might have a better place on Quora instead of an anonymous message board.

If I recall correctly I even saw several well known scientists give well thought out answers to just that question.

  No.550

The only databases you really need are arXiv.org, Library Genesis (currently at libgen.io), and Sci-Hub (currently at sci-hub.bz). They have basically anything you could ever want, so many academics don't even use any of the more traditional or legal resources.

>>548
>organising it (in a pile on your desk i imagine)
Paper is gross and slows you down immensely. Don't print out papers. If you need to write on them then do it on the computer with a stylus.

>>549
>such in-depth advanced topics
Wow, I guess your society really failed you if you think basic research techniques are advanced topics.

  No.551

>>549
You could at least post links then. I was under the impression that Lainchan had a number of university students that could share their experiences here more freely, like in the study methods thread.

  No.552

>>550
>Paper is gross and slows you down immensely. Don't print out papers. If you need to write on them then do it on the computer with a stylus.

I've recently been doing more physics problems with latex, I still use scratch paper but the slowdown that comes with it also means I put more time into my problem setup and rational for the steps.

>>548

This seems accurate, I'm not quite at that level (applying to grad next fall), I'm just really disorganized and have trouble staying on task.

  No.553

>>552
I use a stylus or small whiteboard to do side calculations, but I just write everything in latex now. When you factor in the time it takes to go back and fix mistakes or clean up at the end it's at least as fast as handwriting, if not faster.

  No.555

scholar.google.com helps me a lot.

  No.556

>>549
Honestly Lainchan is just as good as quora for that. I mean don't get me wrong, quora is great too, I use it as well.

  No.608

>>545
For now, I'm a Ph.D. student in CS with a very specific subarea, but I'll do my best to suggest approaches that I think are useful for general scientific inquiry.

>Where and how to find papers

There's a couple of approaches here, based on how specific your knowledge/interest in the field is.

As >>555 mentioned, Google Scholar is an exceptional resource for finding papers for a specific topic (for example, "statistical methods used for scoring coreference agreement"). It's also pretty nice for following citation chains, IIRC.

Otherwise, many people post topics to arXiv.org, tagged with the relevant areas. IIRC, arXiv can email you about new papers in a field you're interested in, or there's a reader that aggregates them, or something like that, so it's helpful for keeping up-to-date as well.

I also suggest finding out what conferences and journals are relevant to your interests and reading what's published in those. Some of these are, of course, paywalled, but it's usually not too hard to find an alternative source (I know plenty of authors pre-publish to arXiv, or will post a non-print copy to their own webpage). I should note that you should look for various signs of the health of a venue: high citation count is one metric, but reading what people in the field think online is also important; basically, look for the conferences and journals that are respected in the field.

>How to read them

Follow these steps, and if at any point it seems too irrelevant, set the paper aside:

1. Read the title
2. Read the abstract
3. Read all the section headings
4. Read the introduction
5. Read the conclusion
6. Read the first sentence of each section
7. Read the first sentence of each paragraph
8. Read the rest of the damn paper.

This is an iterative approach that abuses good writing habits by reading the higher level information first, so you can opt out quickly and not waste your time.

Unfortunately, I don't have any pro strats for digging into the meat of a paper other than to read it and try to work through what they're saying to really understand it. When you find a paper worth digging into, it might take days+ to fully understand it, depending on the depth of the paper and the field.

>How to organize what you have learnt

For organizing papers themselves, Mendeley is a nice tool. You can tag them, put notes, organize them into folders, etc. For organizing the actual knowledge, I've used a variety of tools like OneNote, Evernote, .txt files, etc. Organize it roughly by topic area, keep notes, and keep references to where you learned the thing so that, when in doubt, you can fall back on the original source.

>How to keep up with change

As above, arXiv, conferences, and journals. You could also try to follow science news in the field, or some sort of blog, but I think that sticking to the original sources might be your best bet if you have the bandwidth for it.

>etc.

Let me know if you have any other questions? Also, I'm sure if there are others familiar with research here, they might disagree with what I've said, so feel free to correct me/start a discussion about best practices..

  No.636

File: 1483658340817.png (428.22 KB, 19x200, google searches.gif)

this might help you lainon

  No.637

>>636
>90% of internet users dont know how to use Command + F...
What kind of incomptetent child doesnt even know how to CTRL+F?

  No.638

>>637
Probably a made up statistic.

  No.639

>>638
>>637
It's quite clever, actually.
If you don't know about Ctrl+F, it tells you that it's perfectly okay because 90% of people don't. If you do, you can already feel the elite hacker buzz because you do. Either way, the infographic has been a very nice guy to you, and you should probably show it to everyone.

  No.641

>>608
Are conferences and journals actually relevant for CS/SE?

  No.644

I'm not >>608, but I am another CS Ph.D student here.

>>641
>Are conferences and journals actually relevant for CS/SE?

Absolutely! Conferences and journals are the backbone of academia. If a result is novel and interesting, it will be published in such a venue. The rough hierarchy of prestige goes:

1. Journals. Getting published in a journal is hard, the work is expected to be significantly more in-depth than what you'd get at a conference, and also the presentation is typically much better. It can take over a year from initial submission to final publication, as the paper goes through rounds of feedback and revision.

2. Conferences. These vary in quality, but each field will typically have one or two top conferences. The more selective a conference is, the better it is, roughly. Papers are much shorter than a journal, typically around 12 pages or so. Typically a prestigious conference will have a few more specialised (and less prestigious) workshops and symposiums co-located with it.

3. Workshops and Symposiums. Basically a small conference with a typically very narrow focus, but these can't afford to be very selective because there aren't as many people working in that area! These are often used for publishing ideas which aren't fully worked out yet, to get initial feedback before publishing a more in-depth paper to a conference or journal.

4. University Technical Reports. The lowest of the low. To get a tech report published you just have to convince whoever is responsible for that at your university to upload it somewhere. Generally avoid these, unless they're directly related to a paper (for instance, a more in-depth treatment of something which didn't fit in the paper format) which is published in a more prestigious venue. As these are typically not peer-reviewed, they're not good candidates for citing.

The ACM has a collection of Special Interest Groups which sponsor conferences and journals, which can be a good starting point when you're trying to find something: https://www.acm.org/special-interest-groups

My research is in functional programming, so the ACM SIG is SIGPLAN: http://www.sigplan.org/
The top journal is the Journal of Functional Programming: https://www.cambridge.org/core/journals/journal-of-functional-programming
The top conference is the International Conference on Functional Programming: http://icfpconference.org/
There are a bunch of more specialised workshops/symposiums such as
- the Haskell Implementor's Workshop: https://wiki.haskell.org/HaskellImplementorsWorkshop
- the Haskell Symposium: https://www.haskell.org/haskell-symposium/
- the Scheme and Functional Programming Workshop: http://www.schemeworkshop.org/
- etc

Some other conferences and symposiums I've looked at:
- Systems, Programming, Languages and Applications: Software for Humanity: http://splashcon.org
- Programming Language Design and Implementation: http://conf.researchr.org/series/pldi
- International Conference on Runtime Verification: http://runtime-verification.org/
- International Symposium on Software Testing and Analysis: http://conf.researchr.org/series/issta
- Principles and Practice of Parallel Programming: http://conf.researchr.org/series/PPoPP
- Symposium on Principles of Programming Languages: http://conf.researchr.org/series/POPL

The journal > conference > symposium/workshop order is only a guideline. POPL is, for instance, a "symposium", but it is as highly regarded as some of the big conferences, such as PLDI.

  No.648

>>641
>>608 here, just thought I'd vouch for >>644, who is absolutely correct. I'll add a few more comments, but as I'm still relatively inexperienced, I'm welcoming of corrections/comments on what I'm saying.

I'll touch briefly on the note of conference papers/workshop papers, where you'll find 2--3 types of papers:
- Long papers, which present a relatively substantial result and generally run 8--12 pages.
- Short papers, which can often get away with weaker/preliminary results and run 4--8 pages.
- Proposal papers (these might only be in workshops?) where you submit a topic of interest for the purpose of discussion, which is usually around 2 pages.

As an additional note, you'll often find that workshops are not a standalone thing (although, on very, very rare occasions, they can be). Instead, they co-locate with a larger conference of relevance to the field as a way of not only giving potential attendees in the subfield more of a reason to show up, but also create an opportunity to make more people aware of it. For example, a workshop in my area (narrative) has co-located with CogSci, AAAI, and other major conferences.

I think generally speaking, researchers will publish far more conference papers than journal papers, due to the difficulty in getting a journal paper, but of course all of these things sort of run parallel to each other, so you might see someone publish a few conference papers on a specific thing they're working on, then publish a journal paper to sort of cap it all off.

  No.709

>>644
>>648
Thank you for your detailed answers. I was just wondering because when I was looking for papers about a certain approach to a problem, everything that I found came from a different journal or conference.

  No.710

Is it "safe" to save, say, some papers with notes or textbooks on "cloud services", like Google Drive or Dropbox?
I have some of them and I want to make sure that I will not lose them, so I saved it not only on my external drive, but also on Dropbox.

  No.711

>>709
You're welcome. It'll definitely happen that for a given topic, you'll see papers from multiple journals/conferences. Sometimes it's preference, sometimes it's "we didn't have enough work done to feel comfortable submitting to <x>," sometimes it's "well, we had finished this up in time for this deadline and it seemed like a good place to submit."

There are, of course, more and less prestigious venues, but you have to go pretty far down the venue ladder to get to "this is obviously a bad paper."

>>710
Safe in what way?

Will they get deleted? Probably not.

Will you get busted for having these papers, assuming they're under some copyright? Not if you aren't sharing it with tons of other people.

Is it a good place to store in-progress/unpublished work? I highly doubt that Google/Dropbox/etc. are going to snoop through your files looking for such a thing, and presumably you have some evidence that you're the one working on the paper.

  No.712

>>711
"Safe" = not being deleted/copyright soykaf.
I also use it for in progress work, but no sharing.

Thanks!

  No.713

>>648
>>644
Is LaTeX a must for researchers? How can I learn it?

  No.714

>>712
Yeah, I haven't been hit by it, at least. It's pretty common for people to rehost papers for use within a group or something similar.

I don't think the Elsevier lobby is quite strong enough to have Google content-id that soykaf yet.

>>713
>>648 here, LaTeX is not at all a must. I know of several conferences/journals that provide Microsoft Word templates. I will say, however, that it does make your papers look so nice, and it makes bibliography nice and easy.

I learned LaTeX as an undergrad, when I needed to write a paper in LaTeX: I worked on it with the Wikibooks page for LaTeX open. For most tasks, LaTeX is honestly not that difficult, so you can just dive in; maybe not if you have a deadline in x hours (it would be frustrating), but certainly if you have a decent chunk of time, you can get your first paper looking pleasant, and from there you can build off what you learned (I sometimes look at past papers and copy/paste things that aren't in my working memory).

  No.715

>google scholar
>sci-hub
>libgen
>arxiv

Anything else?

  No.786

>>715
Microsoft has its own academic search engine that lets you search by categories.

  No.787


  No.790

>>713
I would say that it is, but I am a math student. I am surprised to hear that CS people don't all use latex all the time. Having once tried to write something technical (that is, with formulas) in Microsoft Word I can tell you that no matter how you do it it is going to be agonizing. I have no idea how you can write and publish a paper written in that way without it looking like garbage. Plus working with the tex source code is so much cleaner than opening up a WYSIWYG editor when you have a ton on pictures and diagrams.

  No.791

>>713
Oh yeah and like >>714 said, the latex wikibook is great. Other than the documentation that comes with the various packages you use, along with an occasional stackexchange search, it's the only thing you need to learn the language.

  No.792

>>790
Most journals and conferences do require LaTeX, but since universities only want the end-result, rather than the source, undergrads tend to use whatever they're familiar with. Even if it's painful.

  No.803

>>713
Well, it's an industry standard. I strongly suggest you to try it, at least. Remember, typesetting is not easy anyway (eg writing equations is a pain, but there are degrees of pain depending on what platform you choose). LaTeX just always seems to have a some sort of 'hack' to any problem. It is also a markup language system, which appeals to people who like programming. That makes it pretty efficient, but it's also quite monolithic. It takes some time to learn it.

  No.822

>>644
This post was very insightful and educational, especially since I'm also extremely interested in functionnal programming. Thanks.