maling akala

now that i'm in a more relaxed frame of mind [and our grade has been set in ink], the grand adventure of our advandb machine project warrants a more thorough retelling:

the project consists of a 1700-line table of data with sample output, and a 100-line table of data whose output is to be predicted by a program capable of inferring cause and effect rules. this was an obviously daunting task so our group allowed ourselves an extra day just in case we hit snags, meaning we started two days before the deadline.

as i mentioned two posts ago, one group had finished their project ahead of time, and i managed to secure a copy of their source code. foolishly thinking that all we had to do was to create a shell of our own and use their engine to actually run the thing, we didn't bother to doublecheck if our assumptions were correct until 10pm of day one. in hindsight, the file size should have given us an important warning - their project was too complex for us to decipher in a day, let alone three hours [our original estimate].

we immediately relocated to the kitchen table with a ream of bond paper to plan our own system from scratch. the atmosphere was rather grim, since was scarcely enough time to implement the system if we were still on the design phase, not to mention that we were so jittery that the frequent distraction the television provided was more than welcome.

actual coding started at around 2am. around the same time, the aforementioned the day you said goodnight invasion reached its peak, even playing simultaneously on three different channels. equally persistent were jessica simpson's these boots (were made for walking) and brownman revival's cover of the eraserheads' maling akala [more on that later], but neither was as annoying in terms of quantity, even combined. miraculously, by the time the sun rose, we somehow had a user interface and a file input system.

we resumed work at 4pm the next day, and before midnight, we had our data storage system down pat. only two hurdles remained: the algorithm for predicting data, and the algorithm for displaying the results. we used the ide3 approach, which calls for the system to compare input and deduce a decision tree, which in layman's terms is a diagram wherein you start at the top and travel downwards, identifying which branch to take by refering to the input.

for example, the sample data:

red + small + round == apple;
green + small + round == apple;
orange + small + round == orange;
red + big + round == watermelon;
green + big + round == watermelon;

results in the following tree:

1. is it big or small? if big, it is a watermelon; if small, go to 2.
2. is it red, green, or orange? if red or green, it is an apple; if orange, it is an orange.

so if you have big + orange + round == ?, the system tells you that you have a watermelon, despite the unusual color. notice that the 'round' data item is never used, since it's not really helpful in qualifying data.

anyway, back to the story:

we figured that the bulk of the thinking had to go into finding a way for the computer to figure out which data were pertinent to forming the tree and which were not, and although we were right in that regard [we finished encoding the formula by 4am], we suddenly realized that we had no clue how to make the sytem use the tree, once we built it.

the funny thing was that so far, whenever a particularly nasty concept took more than an hour to implement, maling akala would play on some channel or the other and we'd figure it out within the next few minutes. chalk it up to coincidence, a harbringer of divine intervention, or a second wind brought on by blind faith, whatever it was, it was working. when our theory was proved yet again with the final solution to the tree conondrum, we started singing snatches of the song whenever stubborn errors were finally hammered out [usually through some intelligence-insulting means, such as adding a semicolon at the end of a line, or doubling an equal sign].

come monday, we were told that the morning batch of defenses had been cancelled in favor of a wednesday schedule. having selected 11am as our time, we naturally assumed that ours had been among those moved. since we had double-checked the veracity of the schedule change with another faculty member, it was some hours before we realized that some professors have a different definition of the time separating forenoon from after. luckily, we managed to convince "death prophet" that she had failed to make the distinction between morning and whatever the hell she called the hour before noon, and were given another slot.

during our defense, she instructed us to input the entire 1700 line sequence into the system. as we were using a laptop, the strain caused our program to crash, which she then blamed us for. this only served to remind me of the fiasco last year when she asked me to comment out all of the error handlers i'd wired into that subject's project, then deducted points when the system started displaying erroneous results given input she designed, which were well beyond the scope agreed upon for the subject. the difference was that this time, she further deducted 40 points for supposedly incomplete documentation.

but here's the kicker - guess what score we ended up with? 85. that means that if our documents had been complete and our program had run correctly, we would have gotten a total score of 145. how is this possible? because she pulls grades out of thin air. don't think that i'm gloating over this fact. far from it, since some groups i know that deserved high marks received borderline ones.

i would have thrown a grievance suit in her face a long time ago, if not for the risk it would pose to my scholarship. for those of you out there who know who i'm talking about, and don't particularly care about a 0.0 if it's for a good cause, i can provide you with more material, and paxi has pledged to carry out the paperwork.

otherwise, don't make the mistake of having her as a professor. and if you do, don't the mistake of assuring yourself that you'll pass, unless you have at least one chinese groupmate [did i mention that she's a lech?].

marami ang namamatay sa maling akala.

sige ingat.ü 
posted by stani at 10:02 PM