Semester Update

I just finished fellowship application season, which kept me busy writing essays and filling out forms for a few months. Now that I've gotten that over with, I can focus on classes and research again. This semester I'm taking Computer Vision and Mathematical Models of Language. Computer Vision should be useful, as we're covering the full spectrum of vision techniques. However, I am starting to conclude that I have philosophical differences with the professor. His class seems to answer the question "What are some tools I can use to approximate vision?" whereas my question is "What is the underlying mechanism of the visual system and how do we build something functionally equivalent?" The first question is an important practical one, but to the best of my knowledge the brain's visual system is not a hodgepodge of processing blocks (one for texture, one for color, etc.) but a single processing algorithm. The visual cortex is structurally the same throughout the hierarchy, which suggests that any differences between function of visual areas are not innate, but reflect the patterns to which the system has been exposed. I'm not asserting that the brain's visual system is the only solution to the problem, and that we need to replicate it. However, I think it is fair to assert that a patchwork of hacks is no substitute for a sophisticated network and learning algorithm. Mathematical Models of Language is more directly in line with my interests, not surprising given that my advisor teaches the course. So far we're doing a rigorous treatment of hidden Markov models, which I've used many times but never gone through the math at this level of detail.
Recently I've been getting involved in the Graduate Employees Organization recently, which is our union. I'm now the department steward for ECE, which was last filled by someone else in my lab, coincidentally. We're trying to get more engineers involved, which has historically been difficult here. I think it is a combination between the conservativism that is prevalent in engineering, our lack of sociability, and the fact that we get treated relatively well by the university compared to our liberal arts brethren. However, besides my political interest in helping the GEO, I'm hoping to develop a bit more community among grad students in the department. I'm aiming to get some sort of social events going, perhaps a department happy hour or something.
I might have a very exciting announcement about my future research plans soon, but I'm going to refrain from talking about it here until I know if it's really going to come through or not. The suspense must be killing you, but not nearly as much as it is me.

Scientific art

I saw an article at Scientific American recently titled "How do artists portray exoplanets they've never seen?". It got me thinking about an interesting profession: that of the scientific artist. I found online the International Association of Astronomical Artists and Guild of Natural Science Illustrators, the latter seeming to be mostly biology illustration. There are of course medical illustrators as well, whom you should thank every time you see the cross section of a head (or whatever other part) on a poster in the doctor's office. The relative weight given to information content vs. artistic concerns ranges over a wide spectrum. On one end you see things like beautiful but somewhat speculative renderings of planets, meant to catch the eye of the public. On the other end you have things like mathematics visualization where accuracy is the whole point. Either way the goal is presenting information in a way that's more appealing and comprehensible than a list of numbers.
In fact, once I started thinking about this, I realized my friend Matt Hall (with the Advanced Visualization Laboratory at NCSA) did some animated tornado simulations for nova. I can't find a full video of the animation anywhere, but if you go here and watch the video preview, there's a clip of it at the end.

Why support small computer companies?

So you saw in a previous post that I had ordered an ubuntu Dell. Well, I'm typing this on my new laptop and it is not a Dell. When I ordered the Dell they told me it would take a month to ship, which I thought was a bit long, but I'd deal with it, since I really didn't need it until the fall. The month passed, and it got delayed. And then it got delayed again. I did some research on the forums online, and found that many of Dell's laptops have been massively delayed. Not confident that it would ever actually arrive, and having been given a runaround by customer service when trying to figure out what was going on, I got fed up and canceled my order. Instead, I got a laptop from System 76. They are a small company that sells reasonably priced ubuntu systems, and they have a pretty good reputation for both hardware quality and customer service. I paid about the same amount that the Dell would have cost, trading more hard drive and RAM for slightly worse graphics. Nine days later, my laptop arrived. I only wish I had ordered from them in the first place.
As for how it runs, I'm very pleased. The wireless worked out of the box (unlike every other time I have installed linux). I previously used openSUSE, but ubuntu is quite nice. It is so far more user friendly than SUSE. I appreciate that, because I know how to do a lot of the linux dirty work, but that doesn't mean I want to all the time. My only complaint about the computer is the touchpad. It has the tap-to-click feature, which I can't stand. I end up accidentally clicking on things all the time. On my last laptop, it was easy to change by inserting a line into xorg.conf, but that hasn't worked on this computer. I have searched many forums and wikis, tried all the advice for disabling this annoying feature, and nothing works. Of all the computer problems to have, it's not the worst, but I'm going to keep looking for a solution for a while. In any case, I highly recommend System 76 for anyone in the market for a computer, even if they aren't familiar with linux. And I emphatically don't recommend Dell.

Summer is over

And once again the streets of Champaign fill with confused freshman and undergrads eager to return to their partying ways. Its not so bad, though, as I have some interesting classes coming up. My two classes are Speech Processing Fundamentals and Pattern Recognition, both of which will be useful for my research. I also plan on starting on my own research soon, as I spent the summer trying to get other people's programs working on the robots and figuring out how the platform works.
In other news, I've started working part time on an organic farm. So far it has been great, albeit tiring. It's really interesting to learn about how to grow food organically, especially since I'd like to do it for myself someday when I have a house instead of an apartment. An added bonus is snacking on cherry tomatoes when harvesting, and not having to worry about what's been sprayed on them. If you live here in town, check out Blue Moon Organic Farm at the farmer's market on Saturday, I probably picked the vegetables you will buy.

Dude, you're getting ubuntu!

I know I'm not the first to use that joke, but it's true: dude, I'm getting ubuntu! An ubuntu dell, that is. As a recent laptop mishap put me out a computer, I thought I'd give the new pre-installed Linux Dells a try. Last time I bought a custom build from Linux Certified, since I refuse to buy a computer with an expensive operating system that I'd promptly wipe from the hard drive anyways. I found this time that the Dells were a better price for the same specs than any custom build, and I'm expecting to have all my various drivers already there (something I've done myself with all previous linux installs). When it comes in, I'll tell you how it runs. It's nice to see open source hitting the mainstream, albeit slowly.

Peak Hafnium

When you think of human effect on the planet, usually global warming and peak oil come to mind. Legitimate concerns, yes, but what about peak hafnium? I had never thought of it, until I saw this article. Apparently our supply of some metals used in industry and electronics is getting used up quickly. The other metals mentioned are platinum, indium, tantalum, copper, and zinc. Of course this is concerning, as there's not a lot of things harder to find a substitute for than an element.

How I feel about being a Linux user

Need more geekiness? Go to xkcd.com

Deoxyribogooglaic acid

Google has just invested in a personal genomics company, according to the BBC. I'm not expecting to see my searches tailored to my DNA anytime soon, but it looks like Google is definitely interested in the prospect. It makes sense, as their stated mission is to organize the worlds information, that they would be interested in indexing the gene pool as well. Their astounding success at this mission is somewhat disquieting to me, as it probably is to anyone with any sort of libertarian streak. However, they're excellent at what they do, its free, and they're relatively ethical... its hard to complain too much. After all, you're reading this on Blogger (a Google tool).

Summer is here

Hi all, the end of the semester got me bogged down a bit. It's nice though to be reaching summer. In the end, the difference is mainly that between sitting inside reading journal articles and sitting outside reading journal articles, but I'll take that.
I've went to Washington DC a few weeks ago to try and get an EPA grant for the solar lantern project. It's a project I've been working on to provide low cost solar powered LED lanterns to people in unelectrified areas of developing countries. The issue is that many people in the world (about 1 billion) are using kerosene lanterns to light their homes, but kerosene is polluting, lung disease causing, expensive, and a fire hazard. We have the manufacturing cost down to about $9, whereas in India people spend about $20 per year on kerosene. The lantern lasts about 4 years, so while buying it is an initial investment, it saves a lot of money in the long term. When Patrick and Maren went to India, they sold several of the prototypes, so the interest seems to be there.
While we got the phase 1 grant, we didn't get the phase 2 grant. I conjecture, based on the groups who did receive it, that the EPA was more interested in domestic projects. Which is fair, given that they are a U.S. government agency. However, not to be discouraged, we are going forward with the project as a social entrepreneurship project. In other words, we'll run it as a business, so we don't need continual grants and donations, while placing a priority on making something that is affordable and useful for our target communities. We should be getting a website soon, so keep an eye out for that here.
I've decided on a general area for my master's thesis. I want to explore the way knowledge taxonomies are learned. For example, you know that cats and dogs are both animals. And if you saw an animal that didn't match any previous animals you've seen, you'd still recognize it as an animal. So I plan to implement a mechanism in my robot to do this. There will certainly be more details and revisions in the future, but this is the idea I think I'm going to run with.

AI can now be convicted of a crime

Well, not quite. An expert system that helps people with bankruptcy filings has been judged to be practicing law without a license. Of course, the human owner of the service is the one who has really been convicted of a crime. This brings up all sorts of interesting legal questions though. Expert systems are now used in medicine to make diagnoses, although the diagnosis is typically confirmed by a doctor. They have accuracy rates comparable to human doctors, which means pretty good, but not always on the mark. How long until we see a medical expert system sued for malpractice?