Thesis Updates

I don't even know why I am putting this up on the web, I have been putting up various articles up on my blog, old, new, very old ones, but this one I am not even sure why I am writing this.

It so happened that my algorithm that I had thought of for my Masters' Thesis, works and works perfectly, or rather I should say worke"D" perfectly only because there was a bug in the code, now after flushing out the bugs, the thing doesn't seem to work at all!!

I've thought of everything, ranging from further bugs in the code, to simple coding mistakes, and also about floating point inaccuracies, and what all and what not. But I increasingly think that the algorithm itself is a bit flawed, it initially appread so simple to put down, almost unbelievably simple! But as lots of other solutions in computer science, it turned out to be the "elegant, beautiful and wrong" solution.... (OR IS IT ?)...that is still to be decided, I mean it still has to be decided about the goodness or the badness of results it is throwing up, but methinks it is flawed.

Means atleast I now know/concede that it may be the wrong algorithm altogether, atleast it works perfectly for clique intersections, let us give it that much credit. Two perfect cliques intersecting, it catches them pretty well.

Well the intriguing part is that why does it work so perfectly with the bug, for all those who've read this all the way so far, the bug was simply writing '0' inplace of 0 and '1' in place for 1..which translates to writing 48&49 in place of 0&1 (in ASCII codes)...well 48 and 49 worked as pure charms, when it came to that particular graph size (I was using 200 nodes I guess, to test my algo)...the strange thing is, or maybe not that strange - those numbers 48.49 worked perfectly well for that size 200, but bombed for other sizes, like say 70...(for which I used 15&16, and they worked like a charm !!)

So here I am, blogging about this failure of mine, well, not exactly a failure, now that I know there is an approach that won't work, I'm wiser that way...And wiser still, I know a bug that can actually make it work :P

Maybe it's time to ask my guide to allow me to investigate the bug a little, or for him to do so, anyways time is running out fast, I need to have an algorithm, tests on benchmarks, and get everything done by 29th Novemeber, which happens to be a very good day, in order to meet the submission deadline of 30th Novemeber.

PS: The dreaded placements approach :(, they start the next day infact :( and I guess I'll be in Bombay giving the Morgan Stanley interview (hopefully) on the 29th Nov (although the dates haven't arrived yet, but it will be during that time)...

PPS: In the meanwhile, you hope and pray that I get through, I'll do the dirty work of preparing for those interviews !