A Place for Neural Networks?

In my opinion, neural networks are some of the most interesting tools in computer science and other fields. Their applications are enormous, and it would be interesting to see how they can be applied to political issues.

While this sounds far-fetched, there is some research being done in this area. Journal articles are hard to come by in this area, but things like mathematical analyses of hostage situations and viewing terrorist organizations as chaotic structures are topics you will find in international relations journals. Another subject is mathematical sociology.

What about applications to political, or even economic, issues? The interesting thing about neural networks is they can theoretically find patterns that humans may overlook. With systems that have much too many variables for people to actually contemplate (i.e. economic systems, the stock market, civil wars), a complex neural network just may be the answer - or a means to one.

Avian Influenza Networks

The World Health Organization has released dire predictions about the amount of deaths Avian Influenza can cause. While the numbers are extremely high (up to 1.5 billion deaths), there is one interesting factor in all of this - infectious disease networks.

In Linked, Albert-László Barabási explores networks and makes reference to the HIV/AIDS pandemic. The disease spreads through a number of ways, including sexual contact and sharing of needles. As such, one can model its spread through the use of social networks, as the only way one can get the disease is if one comes in contact with someone who has it.

Can the same apply to Avian Influenza? I believe so. Once a highly infectious strain of flu evolves, the first person to be infected will infect a number of people, who will in turn infect more, and so on. This is a network phenomenon, and as in the case of SARS, those who travel a great deal or come in regular contact with a large number of people will spread the disease quickly. Preventing those from getting the disease will slow the spread, though doing so will be difficult.

One interesting characteristic of networks is that they often appear with the "Power Law". Let's say you have a function, f(x) = y. Let x represent the number of people that caused y number of infections per person. For example, if x = 10 and y = 30, it means 10 people caused 30 infections each. Graphing such information will show you that f(x) is a curve exhibiting exponential decrease: one or two people caused a large number of infections, three or four a bit less, than a few dozen even less, and so on.

Yes, this is a mathematical model and yes, if Avian Influenza does as much damage as some predictions say, it will follow this model.

Modelling the Stock Market

"Not really random, but so complex it might as well be," is what usually springs to mind when thinking of the stock market. Finding patterns in the system is something challenging many people, with the incentive being a range of interests in the markets, or simply wanting to earn a heck of a lot of cash.

A recent article published on New Scientist, called "'Zero intelligence' trading closely mimics stock market" says scientists have been able to mimick the London Stock Exchange through the use agents who trade randomly (to put it simply). The article also has a link to a previous one, entitlted "Virtual brokers forecast real stocks", which supports the idea that the stock market, and other complex systems, can be modelled through the use of multi-agent systems.

It is possible to call the stock market chaotic, but saying it is random may be going too far. With thousands of people trading daily, it is difficult to find patterns in the system because there are so many variables and so many unknowns. However, by making a few correct generalizations, it may be possible to learn something. A lot of the lessons in the goal of finding order in the stock market can also be used elsewhere, with economics and even internatonal relations being two areas who stand to benefit.

And even if it is all random, one interesting aspect of "randomness" is that it is self-similar. Random variations look similar whether you look closely and move back and look at the big picture. Paradoxically, maybe this is the "order" we are all so interested in finding?