Neural networks

So yeah, this is the topic I´m working on now. Sounds pretty cool doesn´t it?

In fact, Neural Networks is an optional course in computer engineering. And it is so, optional, so many many will miss it. It´s a pity, though others will consider it a salvation.

What is this stuff? Well, Neural networks is a relatively modern branch derived from AI (Artificial Intelligence yo, dummies). In a sentence: It´s the computational representation of human brain. But this definition is way too far from what a human brain really is, since we do not know anything about human brains yet, or at least not as much to create a 'bionic' copy of it, and make it work out.

Aside, it´s been possible to emulate the behavior of a biological neuron. You know, the electrical pulses between connections, and that chemical substance in between where knowledge is somehow stored. (Remember from the school?). So if a neuron can be emulated, then a whole bunch of them can be emulated creating subnets, nets and so forth. The drawback is the human brain is formed by millions of millions of these little things... Up to now, computers cannot store a representation of an entire human brain network. (Not unless we tile the entire planet with ram memory slots and connect each other to finally reach a mother board).

As far as I know neural networks are used to pattern detection, imagery, function approximation and error detection in industrial processes. Real application are Cancer detection, missile targeting, or character recognition.

Playing around with them under Matlab environment (You engineers should know about the program) tought me they are pretty powerful tools. Basically it´s like raise a little brain as baby, teach it with some kind of data (Images of lung cancer, for example) and then try its response inputting a testing parameter (A new lung cancer image) and see if the network has learned something. It usually does!!

All right, enough engineering for today. Listen to some punk, it will cheer you upright to heaven... ;)