INTRODUCTION
I should begin by explaining what is neural networks[1]. Basically this is a term I have learned in algorithmic composition class, we have been covering most, if not all of different generative systems from starting from Markov chains[2], data mining, going through cellular automata and lately neural nets and genetic algorithms. Although those concepts are mostly discussed in musical terms it is not hard to figure out they are finding their way in every different aspects of science which is connected to generative systems.

The basic and most distinct feature of neural networks between those systems is, the agent in this network is able to learn and improve its skills which makes it similar to us, human brain, if we compare it with different behavior-based systems where the agent behaves according to the predefined rules of its master about its own self, or in response to its environment.

If I try to put this in an example, a basic Walter's tortoise [3] which has two photocells in front of it as sensors is preprogrammed[neurons] to discriminate the inputs it gets according to light intensity and behave according to these set of rules. Basically what lacking here is self-organization. One step later than this is, the neural networks, the agent can be taught to become superior if not perfect. For the above example, it can be said that the agent is not controlled by set of rules but by a learning process. This is rather similar to the way the brain learns to distinguish certain patterns from others. Certain pattern is fed into the input of the agent and it outputs some action according to that, and that action is compared with the target output, thus we can measure the error the agent is making so it can be healed.

OBJECTIVE
So what I want to do with this project is basically to explore how neural networks adapts to its environment, question can we get away from predetermined behaviors, can we search for intelligence somewhere else rather than the interaction between agent and the environment.

Practically I want to build the Walter's tortoise with introducing different algorithms with set of new "neurons" which gives it the 'life". After I build it I want to experiment it in an changing environment which I am going to setup for this purpose. I am thinking to map this navigation and bring it into processing if I could :) So we are going to have the visual data of the experiment as well.

I want this project to go side by side with Introduction to Physical Computing Class Final since what I want to build is a physical agent which interacts with its environment that is changing. I am expecting to find new questions on the way for sure since the whole concept is totally new to me.