What types of problems can neural networks solve? (excluding optical character recognition)

Neural networks seem really c开发者_StackOverflow中文版ool but what types of problems can they solve?

They are good at regognizing pattern, not only characters. So they can tell you is a picture showing a man or a woman or is this picture showing person xy. Note that they never give an exact answer but only probabilities. Also signal patterns like audio signals. For example does sound x sound like sound y and so on.

They can not tell you what is the answer is formula xy, because small changes in the formula can make a complete different answer.

Neural networks can deal with a large number of different problems. For example, in our case, we have used them to successfully reproduce stresses, forces and eigenvalues in loaded parts (for example in finite elements analysis problems). There is a whole world out there in engineering and computer science where they can be applied successfully.

i'd say any problem where an ai needs to learn from previous experiences. e.g. pandoras box or last.fm where the application learns what music the listener wants to listen to by having the listener vote suggestions up or down.