Appropriate Problems for ANN learning:

Conversely as we did for decision trees there it's important to know where ANNs are the right representation scheme for such job. However the following are some characteristics of learning tasks for that artificial neural network are an appropriate representation as: 

1. Moreover the concept as target function to be learned can be characterised in terms of a real-valued function. So that would, there is some translation from the training examples to a set of real numbers then the output from the function is either real-valued or if a categorisation can be mapped to a set of real values. Whether it's important to remember that ANNs are just giant mathematical functions like the data they play around with are numbers, before logical expressions, etc. So this may sound restrictive, hence many learning problems can be expressed in a way such as ANNs can tackle them, in particularly as real numbers contain booleans as true and false mapped to +1 and  -1 or integers and vectors of these data types can also be needed. 

2.  In long training times acceptable. However neural networks normally take a longer time to train than, like for example a decision trees. In fact many factors there including the number of training examples in which the value chosen for the learning rate and the architecture of the network that have an effect on the time utilised to train a network. Thus training times can vary from a few minutes to many hours. 

3. whether it is not vitally important that humans be able to understand exactly how the learned network carries out categorizations. But as we discussed above that ANNs are black boxes and it is difficult for us to get a handle on what its calculations are  well doing. 

4.  Where in utilisation for the actual purpose it was learned for such the evaluation of the target function needs to be quick. Although it may take a long time to learn a network to instance, decide where a vehicle is a tank or bus or car, where once the ANN has been learned, utilising it for the categorization  task is typically very fast. However this may be very important that: if the network was to be utilised in a battle situation so then a quick decision regarding whether the object moving hurriedly towards it is a tank or bus or car or old lady could be vital.  

In fact neural network learning is quite robust to errors in the training data that is just because it isn't trying to learn exact rules for the task, but rather to minimize an error function also.