what is the problem for which neural network is a solution .what is the technique adopted for image processing before invention of neural network.is neural network just a way of programming where processor can be made to learn to solve the process.will neural network get trained even during the testing.can face recognition be done by storing a database of pictures and comparing the test face based on pixel by pixel information with the stored database of faces.what are the ways of implementing face recognition without using neural networks.
Sorry, don’t get "Olivia O"’s answer.
NN are not really my field, but I’ll answer part of it.
The thing about tasks such as face recognition is that they are difficult supervised learning. Supervised means that you know the answers for some inputs & can train on them.
Pixel by pixel is usually a problem for such tasks as when you change the lighting or face angle, the pixels change. What you might like instead is to extract.process the data (pixels?) into some other features, like the separation of eyes, length of nose, scar on cheek, whatever, you don;t really care as long as they are essentially invariant over the things that change & constant for the same faces.
Therefore feature selection is an issue.
NN usually have internal layers between neurons. These layers represent features developed internally during training & you do not necessarily care what the feature is, just that it makes the correct decision.
A NN therefore is a way to implement a trainable feature selector & decision mechanism.
There are likely disagreements between those who like NN & others, but I read it explained once that NN tend to be black box: They will train & achieve a level of performance but you cannot easily determine how they are making the decision.
In contrast support vector machines can easily show you how the decision is made.
NN can be created to represent many traditional adaptive algorithms & hence often in many ways are just the implementation of an algorithm that likely existing in other implementation before NNs.
The other thing about storing databases, is that most images have a LOT of pixels. To store & then calculate & make decisions based on all of them can be a problem: imaging Google sifting millions of images to answer a query. Instead the idea of calculating & storing a smaller set of features is attractive for storage & speed. That is you have reduced the dimensionality of the problem.
Hope this helps. Maybe someone who is an expert can correct my gross simplifications & errors. 
powered by Yahoo Answers