Erik Rauch



I graduated in May from student at Yale, majoring in Computer Science and Math. I'm heading for graduate study at Stanford in mid-September. This summer I'll be working at GMD (the German national institute for computer science), in the adaptive systems group.

You can see my other home page here, with some fun and interesting links.

My vita in hypertext or postscript.

Interests

Nonlinear dynamics and fractals

I'm now working with Benoit Mandelbrot, doing my undergraduate thesis on the fractal properties of boundaries in the Ising model at the critical state. I'm also investigating lacunarity, a measure of the "texture" of a fractal.



Multi-agent systems

I'm intrigued by collective intelligence and the possibilities of achieving AI in a distributed way. A great metaphor for this, or inspiration if you like, is the social insects. An ant colony gracefully shifts resources around, and responds to its environment in a coordinated way, without any sort of central organization. There is no one ant that gives the orders, and each one can sense only its immediate environment; but a swarm of them exhibits global cooperative behavior, and the species has been incredibly successful. Maybe this very lack of central command is a source of strength, and in some sense the functionality of the whole is more than the sum of its parts.

At the Center for Nonlinear Studies at Los Alamos and the Santa Fe Institute, I worked on a project started by physicist Mark Millonas to better understand this kind of mass action by creating a model containing what we believe to be the minimum components necessary for swarm dynamics and analyzing it. Our conclusions are about the functionality of the swarm. In particular, it has the ability not only to form stable patterns of traffic, but to shift them in response to a small outside stimulus. We found that the swarm has this "information-magnifying" ability when its behavior is just below a second-order phase transition separating ordered from disordered behavior. If you're interested, I am the first author of a paper in Physics Letters A that you can read in hypertext format or postscript (with figures).

(If you want to hear more, and happen to be going to the March conference of the American Physical Society, I'll be giving a talk on it.)

Evolutionary computation

I've also worked on genetic algorithms. In 1993 I wrote PMPGA, a parallel GA in C++ and Linda, as part of a computational biology project. The GA is being used to optimize 60 floating-point variables in simulating a mathematical model of the development of the fruit-fly embryo.


Related links

  • Adaptation, Co-evolution and Learning in Multiagent Systems
  • Dynamical Systems and Learning at Brown
  • Information Mechanics Group (Physics of Computation) at MIT
  • American Physical Society March Meeting
  • Fifth Annual Conference on Evolutionary Programming
  • Neural Nets and adaptive comp. at LANL
  • ALife Bibliography

    Local resources

  • Computational Ecology at Yale