So, I am sitting here in the BMT Unit, and I decided to listen to some music. So, I plugged the speakers to this computer into my PDA and turned the music on... my initial concern was that the music might bother some of the patients, but I quickly learned that whatever residual hearing these guys may have had before they were admitted to this unit... we kill it all off with our chemo.

I have been fiddling around with a program to try and interpret images, and have been met with only marginal luck since I have been trying a brute force and admittedly unsophisticated method. So, I decided that it was time to learn a little bit more about computation theories... in particular, how to make an adaptive program. You see, the main problem that I have been having is that tolerance thresholds for significance vary from one region of an image to another, and I didn't have a way to allow the progam to know when it was on the wrong track. I could get it to pretty accurately figure out what areas represent the majority of the body wall, but after that, the reality of finite resolution hits me square in the face.
Fortunately, I happen to pick up a book on AI theory, which is not nearly as complicated as one might think. A lot of the algorithms are the same as those used in optimization theory, such as simulated annealing and DNA mutation based algorithms. What struck me as brilliant is the way that artificial neural networks are set up. I think the brilliance of them is that the fundamental idea underlying the networks is simple in its concept, regardless of how complicated the final calcuations may turn out to be. First things first though, I need to learn more about the particulars of their implimentation before I start reworking the entire program to incorporate it.
Another topic that I found amazingly useful is fuzzy logic which allows a program to essentially make imprecise decisions based on a given set of inputs in order to compensate for real world variabililty in measurements. I have a feeling that when it comes down to the nuts and bolts of the program, I am going to have to incorporate some of these algorithms into the program as well since most low resolution images have a great deal of background noise that make determining the significance of true signals difficult.

So, I had a day off today... I have been thinking about putting a computer in the living room for a while now, so i finally got around to doing it. Since it a pretty packed room to begin with, I couldn't really have my full linux box in here... also, during my move down here to San Antonio, i managed to bust the screen to my laptop. Logically, I decided to load red hat onto my laptop and just hook it up to a wall mounted flat panel (I got a great deal on an open item 17in HP at best buy). I have to say, I have undertaken a lot of project in my life that haven't turned out so well, but this one is actually turning out exactly like I expected it. If I get around to it I will post a pic of my setup.