Lets say you have a set of images and a set of image features, the image features are a huge set of various statistics performed on the image. You are trying to use image features to classify these images, and these image features may be inter-related. We are essentially trying to produce a graph that shows that these features are dependant (positively or negatively) with one another.
Lets say there are 5 image types, and 1000 features.
Lets say you had enough data to calculate how each feature was related to the type of image. Say, the image analysis algorithm is represented by a formula R(f) for f=(1 ... 1000), but features can be corelated. What strategies are there to produce the neural network graph for problem? See the following diagram:
Is there a particular mathematical field that deals with this? What is it called? Combinatronics?Code:http://upload.wikimedia.org/wikipedia/commons/e/e4/Artificial_neural_network.svg
Hope this makes some sense...