Comments on: Thesis Status Update http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/ I will not fix your computer. Tue, 04 Aug 2020 22:34:33 +0000 hourly 1 https://wordpress.org/?v=4.7.26 By: Luke http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4228 Tue, 24 Apr 2007 01:22:41 +0000 http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4228

MS in CS :)

It goes like this:

BS = Bull Shit
MS = More Shit
PHD = You still don’t know shit, do ya, doc?

Seriously, I want to get a PHD one day, but I think I need to get out and work in the industry for a bit now. Otherwise I will be overqualified to do anything.

I mean, I could teach and do research. I guess I wouldn’t mind that at all.

Then again, I’m not sure if I’m cut out to be a full time research prof. :/

Reply  |  Quote
]]>
By: Matt` http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4226 Mon, 23 Apr 2007 19:25:15 +0000 http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4226

note to self: processing hyperspectral images is really complicated

Well done :)

oh.. and I always forget, which qualification does writing a thesis get you?

Reply  |  Quote
]]>
By: Luke http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4223 Mon, 23 Apr 2007 13:46:04 +0000 http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4223

In essence it is a multi-threaded implementation of two Nonnegative Matrix Factorization algorithms (basic iterative version, and a projected gradient version). With the first one I was able to cut the execution time in half with 4 CPU’s (or 4 times with 8).

We’re using these to process hyperspectral images – and there is no standardized data format for these. Essentially every tool writes the data in their own way. So a pixel can be a byte, a 32 bit integer, 64 bit integer or a float either 32 or 64 bits long. All the non-floating point types can be signed or unsigned. Oh, and some ship in big endian, while others in little endian.

Then there’s the method of encoding hyperspectral bands in the file – some people just write each band in sequence (BSQ), some interleave them by pixel (BIP) or by scan-line (BIL).

So I have bunch of tools in there to read, write and handle this data, display them on the screen and etc. There is also a mostly functional GUI in there that let’s you do all of that.

Of course there are bunch of features that should still be added, and some code is not finished – for example, I only have support for the data types that were available to me. So it should have another 1000 lines or more to even approach being completely finished.

But the most important part of my work, which this thesis is actually about is the parallelization code, which is complete, finished, and extensively. :)

Reply  |  Quote
]]>
By: Fr3d http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4222 Mon, 23 Apr 2007 07:59:48 +0000 http://www.terminally-incoherent.com/blog/2007/04/23/thesis-status-update/#comment-4222

Good luck :)

May I ask what the code/program does? :p

Reply  |  Quote
]]>