Machine learning, in its broadest sense, is a series of methods to recognize and exploit patterns in data. The name comes from the goal of trying to automate (via machines) the process that humans have used to observe the world around them and draw conclusions (i.e. learn) from those observations. Although all practical work in the machine-learning field is done through computer programming, the concepts are independent of programming knowledge and instead rely on a mathematical basis. This overview will look only at the conceptual and mathematical side of the field, with little mention of the programming or practical applications.

There are a multitude of algorithms that are grouped within the general category of machine learning. Depending on the type of information available, as well as the goal of a problem, many techniques will not work well or simply be impossible to apply. The key to learning different algorithms is to know in which situation each functions best. In many situations, there is some sample data from a system, and the goal is to interpret this data to define the system or to predict the behavior of new situations. These techniques will be examined later in the overview. Initially, problems will not provide sample data, but instead define a problem according to some constraints; the goal will be to find an optimal solution given the constraints.

Machine Learning: The History of Automating Computers to Observe and Analyze Data looks at the attempts to develop machine learning, from successes to failures. You will learn about machine learning like never before.