Appropriate Problems for ANN learning - artificial intelligence- 

As we did for decision trees, it is essential to know when ANNs are the correct representation scheme for the task. The following are some characteristics of learning tasks for which artificial neural networks are a suitable representation:

1.   The concept (target function) to be learned may be characterized in terms of a real-valued function. There is some translation from the training examples to a set of real numbers and the output from the function is real-valued or (if a categorization) may be mapped to a set of real values. It is essential to remember that ANNs are only giant mathematical functions, so the data they play around with are numbers, rather than logical expressions, etc. it can sound restricted, but several learning problems may be expressed in a way that ANNs can deal with them, especially as real numbers contain Booleans (true and false mapped to +1 and -1), integers, and vectors of these data types may also be used.

2. Long training times are acceptable. Generally neural networks take a longer time to train

Then,  for instance, decision trees. Various factors, including the number of training examples, the value selected for the learning rate and the architecture of the network, have an effect on the time required to train a network. Training times may be varying from a few minutes to many hours.

3. It is not vitally essential that humans be able to understand accurately how the learned network carries out categorizations. As we look above, ANNs are black boxes and it is hard for us to get a handle on what its calculations are doing.

4. When using for the definite purpose it was learned for, the evaluation of the target

function needs to be fast. While it can take a long time to learn a network to, for example, decide whether a vehicle is a bus, car or tank, once the ANN has been learned, by using it for the categorization task is usually very fast. This can be very essential: if the network was to be used in a clash situation, then a quick decision regarding whether the object moving hurriedly towards it is a bus, tank, car or old lady could be vital.

In addition, in the training data neural network learning is quite robust to errors, because it is not trying to learn exact rules for job, but rather to minimize an error function.