Optimizing a non-linear problem using traditional newtonian methods is often infeasible. Then you may try a grid-type of search, trying to exhaust all of search space. If your space is too big, you have to try non-exhaustive methods. One interesting method is known as the Particle Swarm-algorithm. Read an introduction.
See how an American career-handbook describes the working life of Economists and Marketing Research Analysts. 3 pages (pdf) of interesting reading. I find the section on how some PhD's teach at university while doing part-time consulting especially inspiring.
by Niels on Thursday, November 15, 2001
Sometimes one gets the impression, that no examples exists in which ANNs outperform traditional methods in a practical setting. In Alon, Qi and Sadowski (2001) "Forecasting aggregate retail sales: a comparison of artificial neural networks and traditional methods" we are given an example.
by Niels on Saturday, November 10, 2001