Dec/090
Focused Papers
The goal of most research papers is to focus on a particular objective or hypothesis (survey papers being the major, valid exception), but most papers have lots of extraneous details coming from the exact system and configuration under study.
After discussing the goal of boiling down a problem to its minimal required complexity with Phil Bones and Allan McInnes, I was reminded of the following paper comparing selection mechanism in GA. The point of the paper contains some survey-like discussion, but I admire the simplicity of the system they study. The system choice makes the selection behavior easy to illustrate without extra complexity.
A Game-Theoretic Investigation of Selection Methods Used in Evolutionary Algorithms -- Ficici, Melnik, and Pollack
Oct/090
Kids doing research: Blackawton bees
Beau Lotto and 25 under 8-years-old collaborators produce new results relating to the behaviour of bees. Beau Lotto and his eponymous research lab do research on perception, mixing seeing and hearing, and adaptive things (life and mechanical approximations of life).
It's cool beyond words, and it's got me thinking about education, how it's done and how it could be improved. And it would be totally awesome if a bunch of kids got published before reaching the double digits!
Peter
Sep/094
Iterated Function Systems
Fractal Flame iterated function systems are fascinating. I've always though a simulated robot that was the iterated particle would make for a good RL problem. Basically, the robot would be iterated in space by the IFS and it would need to learn to navigate to target locations using small forces (think of it as a waterbug in a swirling pool).
The source in C++ (requires libgd). Here's an example of an IFS generated from a billion samples.
Aug/093
How to do research in the MIT AI Lab
How to do research in the MIT AI Lab is an old, but by no means outdated, compilation of tips for postgraduate researchers. It was assembled by various members of the AI Lab based on their own experiences. There's a wealth of information on writing, giving talks, picking a research topic, doing research, and dealing with the emotional rollercoaster that is research. Well worth a read if you're a postgrad. Probably also worth a read if you're supervising postgrads, since it has some great advice to pass on to students.
The link above is an html version hosted by Indiana University. A PDF of the original AI Lab working paper can be found here.
Aug/092
Research update: Predicting the effects of turbulence.
Here are some interesting images of the Jovian moons from my last field trip to Mt. John Observatory, near Lake Tekapo in New Zealand. Note, I have updated this since my earlier post and qualified several parameters.
Aug/091
Empirical Methods in Machine Learning & Data Mining
Interesting page from a 2003 Machine Learning post-grad course at Cornell:
http://www.cs.cornell.edu/courses/cs578/2003fa/
The lecture notes are fairly good and the page contains links to other good resources.
Aug/092
Paper Writing Steps
Many students have problems getting started on or making progress with paper (or thesis) writing. The typical process that is usually tried is:
- Outline content of paper
- Write each section of the paper (usually in order)
- Add figures/tables/graphs
- Add references
- Write abstract
- Figure out a title
I claim this is almost the opposite of the best way to write about your project/research. Let's consider what you have to decide as you write a paper (here I'm assuming we're trying to write a good paper as well).