Feb 28, 2012
Posted in: Academics & Research
In the hassle of getting my life and graduation requirements together, I almost forgot to take advantage of my (last :'/) biannual free post: what am I doing this semester?
MIT has many urban horror stories about students struggling to graduate on time for the dumbest possible reasons: taking all of their GIRs second semester senior year because they didn't spread them out, realizing they forgot about the swim requirement a day before graduation, or belatedly discovering that one humanities class didn't count toward the requirement they thought it did. One alum I worked for told me that he convinced the EECS department to let him take his GIRs during his MEng, because he spent all of his undergrad years taking fancy high-level classes. Sometimes, students just punish themselves intentionally by taking eight classes a semester so they can graduate in two years, or get nonexistent triple majors, or...for fun. (There's no credit limit for freshmen with sophomore standing or upperclassmen; the only bound is your sanity.)
Here's the anticlimax: I am not one of those people. (Unless something has gone horribly wrong.) My courseload:
6.851: (Grad) Advanced Data Structures, taught by Erik Demaine, who became an MIT professor at age 20. He frequently cites his own papers in lecture, and we have optional, weekly open problem solving sessions, which is kind of intimidating. I say that a lot. I mean, other professors have (jokingly?) put open problems as extra credit on problem sets, so it's nice that ours are just completely for fun.
This class is my official substitution for 6.006, which I need to complete my major, so the downer is that I can't drop it when catastrophe strikes. It's been fun so far, though. Seemingly half the CS/math majors I know are in this class, including a '14 down my line of succession from my high school robotics club, and an ex-IOI freshman on my hall. Small world!
21M.380: Recording Techniques and Audio Production is one of those very cool, random, ultra-pragmatic classes that pops up at MIT every once in a while. The assignments so far have consisted of close, detailed listening to songs of our choice, and analyses on how they were mixed. The lectures are refreshingly less theoretical than the average MIT class; instead of generalized, abstracted equations, the lecturer shows us interesting psychoacoustics trivia and heuristics that are useful to know off the top of your head when doing audio production; for example:
- a bass drum's frequency is usually around 60Hz, so its wavelength is about 18 feet long
- human ears start perceiving time offsets as reverberation after 20-30ms, and it takes sound about 0.89ms to travel a foot, so do the math before recording in big rooms
- since the ability to understand human vocalizations was obviously beneficial when we developed speech, human ears are heavily biased toward frequencies representative of human speech and will in fact perceive them as significantly louder than frequencies outside that range
I keep getting distracted by curiosity about various physics-based/biological/statistical generalizations of what we learn. Mostly, the teacher gives us demos (listening and generating lots of noises and signals in different contexts, through filters, et cetera) and enough theory to internalize them.
Incidentally, this class completes my four-subject humanities concentration in music.
Every engineering major has a thesis, although most use a heavy-duty lab class in lieu of an independent project / research paper; Course VI is one of the few with an open-ended thesis project. I'm doing some sentiment analysis with the Media Lab's Digital Intuition group, which does a lot of neat natural language processing that I didn't really know was possible, using a project called ConceptNet that they've open-sourced and collaborated on with universities in several countries. It has a basic web interface and an API that lets you access a massive semantic graph (hypergraph, rather) about words and concepts in several languages. Since everything's open-source, you can go play with it if you're curious.
At only 30 units, I'm light-loading this semester, which means I get a tuition discount! The classes listed above are all I need to graduate, so I'm casually following along (without enrolling) in MAS.S60: Practical Natural Language Processing, incidentally taught by my thesis advisor. In junior fall, I took a similar but more rigorous and theory-heavy class, 6.864: (Grad) Advanced Natural Language Processing. I'm poking through the MAS.S60 material because it has coding labs in lieu of enormous mathy problem sets, and while 6.864 was fun, I didn't get to actually implement code until the final project.
I'm also taking an introductory animation workshop through the Student Art Association, which organizes extracurricular art classes that serve as practical alternatives to the tempting logistical nightmare of crossregistering at and commuting to MassArt. Today, we jumped right in and started making cyclic hand-drawn animations.
is this not enough for you
I have five sessions of job interviews in the next two weeks. Three of them span half a day. No comment.
In case it wasn't obvious, the underlined text in this post contains useful alt-text.