Wikipedia:Reference desk/Archives/Computing/2012 January 2

= January 2 =

Research design for IT PhDs
I'm interested in the sorts of research design used in IT for PhD students. You have to contribute something new, but say your research involves coding something - how do you evaluate it to show it is new? Do you show that no one has ever been able to write such a program before that does such and such? Or do you show that it does the same thing as another program, but using less disk space, or else what? I'm enrolled in a PhD in IT, but having trouble refining the topic, as I know how to program alright, but I haven't had much formal training in the discipline as it works in academia. Thanks in advance, IBE (talk) 10:47, 2 January 2012 (UTC)


 * In general the way you show something is new, in academia, is to show that you are aware of the state of the art and its deficiencies and limits, and that your work is just beyond those. It's not unlike how it works on patents — "My product, product X, is meant to do A, B, and C. Products Y and Z can do limited versions of A and maybe C, but are deficient both because they don't combine these elements together, but also totally omit B, which is obviously a good idea." As for the exact form, or the exact state of the art you should be showing mastery of, that's what advisors and peers are supposed to be helping you with. Your advisor should be helping you figure out what is the level of originality you need to achieve for them to be happy (and thus sign off on the work) — whether you have to re-invent the wheel, or whether you just need to optimize it a little bit. --Mr.98 (talk) 14:24, 2 January 2012 (UTC)


 * If computer science counts as IT, the answer is that coding (as traditionally understood) isn't really a core part of the research process. It's an excellent idea to write code, as a sanity check and as a way to evaluate the practicality of one's ideas, but the products that people produce are generally more mathematical constructs than code (though it's possible to be both). You evaluate these constructs by the features they support, their simplicity, and sometimes even by their performance (as measured in big-O notation). Paul (Stansifer) 17:16, 2 January 2012 (UTC)


 * It would also be a good idea to look at some previously defended dissertations at the same field and, if some are being defended nearby, to observe a defence or two. --Martynas Patasius (talk) 19:52, 2 January 2012 (UTC)


 * Code is not new. It never ever is new. It is like claiming a sentence is new. It just a collection of words other people have been using for many generations before you. What is (or may be) new is the algorithm that the code implements, just as a poem made with words may be new. So, the thesis is on the algorithm, not the code. The following is how I developed my PhD thesis, which was an algorithm, which I also implemented with code in two programming languages:
 * First, I had my idea. That was useless. Everyone has ideas. I had to check with experts (professors and researchers) to figure out what fields of research I was touching on. There were a lot of them. I had to read a lot of books and research papers and then I was interviewed by my PhD committee to ensure that I was competent in all of the fields my algorithm was based on.
 * Next, I had to write a survey paper. This was designed to do three things. It had to show that I had a complete understanding of all the algorithms and theories I was planning to use. It had to explain to anyone else who was going to read the paper how those algorithms and theories related to my goal. Finally, it had to explain exactly why I believe that my algorithm is worth working on. I had another interview with my PhD committee to prove that I knew my stuff and my algorithm was novel and new. That was a very hard interview because everyone on the committee has the goal of finding any fault of any kind in my plan. It is hard to defend against a large group that are all against you.
 * Finally, I had to write my thesis. My survey paper became the background (much shortened). Then, I explained the algorithm in detail. Then, I explained how it was implemented. Then, I explained how it would be tested - and more importantly, why those tests are the best and most valid tests. Just to be complete, I showed the results of the test - which must be good or the thesis wouldn't be published. So, that is a very short section. In the end, I explained different ways in which this algorithm would be beneficial to other areas of research. The last step is publishing that paper and defending it against attacks from the PhD committee.
 * Hopefully that helps. What you really need to do is discuss this with your committee. They know what they want and they could care less what my committee wanted. -- k a i n a w &trade; 20:18, 2 January 2012 (UTC)

Thanks - these replies help a great deal. I know to ask my supervisors, but they are on holidays, and I also can't sit there chatting about how the discipline works in general in academia - and I always get good replies here. In fact, I come to Wikipedia when I'm at my wits' end, and it at least restores my sanity, so thankyou. Note that the stuff about my personal situation wasn't the exact question, just background to help focus the answers - the questions were as listed, and about the general situation, so you've given a lot of help. Any more info is much appreciated. IBE (talk) 08:45, 3 January 2012 (UTC)

rsync not working as expected
Hi, I was trying to consolidate several backup generations using rsync's hardlink ("--link-dest=") feature, to save some space.

What I have: /oldbackupdir /oldbackupdir/backup0 /oldbackupdir/backup1

What I am trying to create: /newbackupdir /newbackupdir/backup0 /newbackupdir/backup1

With the difference that files that are identical between backup0 and backup1 will be hard-linked.

Step 1: rsync -aPv /oldbackupdir/backup0/ /newbackupdir/backup0/

Step 2, try #1: rsync -aPv --link-dest=/newbackupdir/backup0/ /oldbackupdir/backup1/ /newbackupdir/backup1/ And this is where everything breaks: I'm not getting any hardlinks, even though I know some files are identical (same md5sum).

Step 2, try #2: rm -rf /newbackupdir/backup1/ # Clean everything up for a new attempt rsync -aPv --link-dest=../backup0/ /oldbackupdir/backup1/ /newbackupdir/backup1/ I read somewhere that one should use relative path names for link-dest, but it doesn't help, either.

Step 2, try #3+4: Repeat the above two steps with rsync -aPvc --link-dest=, as the file creation/modification times might be different between the backups, but the md5sums are the same for identical files between backup generations, and -c should switch from time/date-dependent to md5sum-dependent detection.

Same result though, no hardlinks. :-(

Step 2, try #5: rm -rf /newbackupdir/backup1/ # Clean everything up for a new attempt rsync -aPv --link-dest=../backup0/ /oldbackupdir/backup0/ /newbackupdir/backup1/ Intentionally copying the wrong (backup0 instead of backup1) source to the backup1 directory does create hardlinks.

What the heck am I doing wrong? -- 188.105.112.97 (talk) 15:48, 2 January 2012 (UTC)

Green and blue look fine on my laptop.
The red, however, isn't red enough, and counter-intuitively for me resulting in a slightly purplish look where it should be white or gray, especially gray. Store's run out of the model, but I can still return it Great price though, and a rare matte. What should I do? 66.108.223.179 (talk) 20:39, 2 January 2012 (UTC)


 * Use a color management system to calibrate individual gains for red, blue, and green channels. If your laptop runs Windows, use the Windows Color System tool, already installed on your system.  Additional technical documentation is available from Microsoft's WCS page.  If you are running a different type of system, let us know and we can help you find another solution.  Nimur (talk) 21:15, 2 January 2012 (UTC)


 * If everything looks purplish, that likely means you don't have enough green (or have too much blue and red). So, try adjusting those levels accordingly.  Under Windows, Start + Control Panel + Display may have some settings for color adjustments, depending on the graphics card, driver, and OS level.  There may also be an icon with a picture of the monitor, down by the volume control, etc., which might allow for color adjustments. StuRat (talk) 21:45, 2 January 2012 (UTC)


 * If no adjustments produce the right "red" color, it may be a hardware issue relating to your graphics card or monitor. It's often a tell-tale sign that something's not working right if either the red, green, or blue suddenly short out. I've had cables where the red was totally defunct and only blues and greens were coming through. --Mr.98 (talk) 01:30, 3 January 2012 (UTC)

If the problem is with an external monitor, do a similar wiggling thing for your VGA/HDMI cable. TheGrimme (talk) 16:07, 3 January 2012 (UTC)
 * If the problem is with your laptop screen and a particular primary color isn't coming through, I would assume the problem is with a short in the internal video card or a lose connection. If you do some Googling for your notebook model with the keyword "Disassembly" you should be able to find a guide for accessing the video cable. Try wiggling it around while grounded with the computer on to see if the color comes back.