Roughly speaking, machine learning comprises a series of methods, that enable computers to adopt certain desired behaviours based on the training of given examples instead of programming the computer with hard rules. This approach is beneficial especially for complex problems, because clear rules or equations are rarely available or known in such cases.
Beside the realization of (self-)adaptive systems, machine learning methods can be used to uncover a system's underlying rules to describe its behaviour a posteriori. In this context machine learning methods are the basics for data mining.
In addition to artificial neural networks, decision trees and Support Vector Machines, for example, are also part of the various machine learning methods.