No correlation! Oh no!
So the hope of the tests that I have written was that there would be a strong correlation between the number of times detected of faults from my tests and from Yiwens of the same size... but that is not the case. The graphs do look similar when all the faults are looked at, but when you look at faults that are least detected, there is really no correlation. In some cases my tests are better, in some cases Yiwens are better. Of course this is assuming that hers is the optimal solution... which it might not be. Not sure where to go from here. I attempted to look at larger amounts of faults when forming my list of vectors... but that didn't work either. Even when I include all of the faults in my calculations, there is only a mild correlation between hers and mine... which is rather odd considering that I am supposedly using about the same algorithm as her. Sort of stuck there, was hoping there would be more correlation... And the thing is, sometimes mine detects more critical faults, and sometimes hers does, and this does not have anything to do with the size of the test set. Huh? Is it all random?
Now Prof Dworak wants me to look at bridge faults, and see if those have more of correlations. I was looking at only stuckat faults before. This involves using FastScan... which I have been avoiding as much as possible. :( I'm procrastinating on that a little by organizing my files. I have been saving everything in the same directory and it looks a little scary... and I havent left any comment on any of my scripts. So did that for the second half of the day today, I think its productive, in its own way. :) My file system looks very nice now. To work tomorrow!