Research Proposal for KR class

This seems to be a sort of fundamental problem - Larry asks his students to come up with some kind of a research proposal (RP) similar to those submitted to NSF (National Science Foundation) in order to get financial support. The reason he's doing this is to get us familiar with asking for money :) Everybody needs money. I need money to study but unfortunately it's not going to fall into my pockets. NSF is one of potetntial financial cows to dry off, so it'd be stupid to not explore this opportunity in the future.
Surely we're not going to submit what will be written in next few days - it'll be rubbish. But! It can give us good ideas of what it takes and what might help you to succeed down the road.
My personal problem is that I do not know what my area would be. Larry has suggested some research in the area of managing impreciseness when building (learning) ontologies for Semantic Web. This sounds pretty scientific but I have a little understanding of the current "state-of-art" - what research is underway, which directions are promising which are not, which attempts have been futile, who made any breakthroughs on what and when, etc, etc, etc... Knowing this stuff requires lots of readings - technical papers, journals, books, conferences proceedings... I simple haven't had a chance to go through that (hum, I could use my XMas break to improve on that) and this's what stops me from progressing.
However it's not a good time to panic (in fact, i don't know any time suitable to fall into a depression). I used Thanskgiving break (4 days) to surf web sources looking for any writings concerned with the area. I did find quite a few of them (43Mb of tough technical stuff). The main idea I got out of that is the possibility to employ Fuzzy Sets (and Logic) techniques for representing inherently vague information. Many people believe this is the most natural way of dealing with categories commonly used by humans that are not clear, formal creatures (unfortunately!). Therefore if we're going to mine textual sources (like most of Web's content) in order to build conceptualizations (ontologies) they are also going to be imprecise. Not only because we're uncertain in their borders (although we are) but also because the borders can't be precise even if we have full information about the Universe. That's why we're looking at Fuzzy stuff not at any means of dealing with uncertainty (Bayesian networks, Dempster-Shafer, etc).
So at least I know now what I'm going to read and where (very roughly but still) I'm going to apply my modest brain.

Final DB exam

Very last and comprehensive DB exam is going to be held on Thursday.. I really need above 90% on tha given my awful perofrmance of first two attemps. But being pinned down doesn't give me any confidence, I don't see how I'm better prepared than I was 4-5 weeks ago. Still having troubles with fairly vague and imprecise questions that allow multiple answers and God knows which one is going to be considered correct (or Larry might know as well). Nonetheles let's stop crying and do what can be done here...

"Automata.." looks good now!

Suprisingly enough Dr. Schmidt indeed gave me 95% on the recent exam. That's really unexpected given I failed on the last problem but he took off only 5 points.. Anyhow - it's really nice because now I'm clearly on the right track to A (also got 19 out of 20 on the last hwrk). Way to go!

Bad week

Well... it was not a great week in terms of grades received. "Shit happens" (c)...
1. Databases. This is the most disappointing course. 73 points was my grade on the second exam which is just miserable.. awful. The funny thing is that I got 15 (!!!) points taken off for not underlying primary keys on normalization problems. Those were hardest and only a few people solved them.. i did. But no keys means zero points although the decompositions were impeccable. Apart of that I made 2 real mistakes and 2 questions I'll probably ask to regrade although it's not gonna change a lot. I'm in trouble in this class, 3rd exam will be crucial for me.
2. Automata. Midterm was held on Tuesday and was probably graded by Prof. Schmidt (not by the TA). Results were posted on Web just couple of hours ago and my score is unbelievable 95/100 (average is 75 including undergraduates who get full credit for 1 skipped problem). I suspect it's a mistake. I couldn't earn 95/100, I failed on 1 problem and 2 other solutions were fairly dubious. So, no joy so far...
3. AI. We had 2 quizes (small tests) and some homeworks. I made mistakes on both quizes (quite stupid) but got 9/10 for first one and 100% result on homeworks. I think Dr.Bhatnagar is too loyal when grading...
4. Knowledge Representation. We (I and Julia) were giving a talk today on "Structured Descriptions" which is a branch of logic used to automate the process of organizing logical concepts into hierarchical taxonomies. The talk went quite well (apart of some purely technical issues), I got not a bad reference from Larry (93/100, understanding - excellent, presenting - good). So KR is fine so far but later on we'll have another talks (on ontologies), an oral exam and some sort of Research Proposal. Which must be good for me given my terrible performance in DB class...

Automata's exam coming

Now it came to the real exam. Big test, short time, long calculations and no questions to skip (this is only possible for undergrads). I'm especially concerned with long transformation of an NFA to the minimal DFA - the kind of problems where extremely easy to make a mistake. Another popular sort of problems is Pumping Lemma applications, problems which require a lot of experience to quickly take right strategy but which don't usually involve tedious calculations.
To sum up, I'm going to sacrifice my WE in behalf of this crap... Which is useful anyway as long as Automata's questions will be in DQE.

DB exam has gone. Bad expectations

Well, I do not expect anything good from it. I've skipped 4 questions (as required) and all of them were multiple choises questions. The fact is that for them it's possible to receive partial credit whereas other (like more hard normalization problems) are graded on right/wrong basis. Which means that I ran into a lot of risk not skipping them... But I still hope that Larry's gonna give for them more credit, just because they're hard. If that's not the case I'm in trouble because I spent a lot of time on them and not very confident in solutions.
So, will see next week, but anything above 80% would be a real fortune...

Automata's homework

38 out of 44.. brr.. everything under 90% is terrible. Especially when getting points taken off for nothing.
However, I must be wrong because Dr. Schmidt is an extremely precise and accurate person (he's German, btw). But I really don't see where I'm mistaken in the following exercise: "prove that the language L={www(w)r; w belongs to {a,b}*} isn't regular." w(r) is an inverted w.
The way it's expected to be proved is applying the Pumping Lemma. However it might be tedious to do so directly, that's why I've taken another way. Let's define some terms first: tail(L) is a language such that: {w: xw belongs to L for some x - any sequence of terminals}. It has been proved that if L is regular, tail(L) is regular too (by construction).
It has also been proved (by Schmidt) that w(w)r is not regular (using Pumping Lemma). But w(w)r is a language from tail(L)! Which would be regular provided L is regular, right? Then L must be irregular... This is my way of proving this.
However, TA has taken 2 points off arguing that Pumping Lemma is expected. Any ideas where I'm mistaken?