4.2 Examples

This subsection gives some simple examples where perceptrons are used to classify data from different classes. Remember that the perceptron is a very simple type of classifier and in most real problems you might need more advanced models.

The first two examples use two-dimensional input data. This is nice since the data and the classifier can be depicted graphically. In the first example there are only two classes, which can be separated using a perceptron with one output. The second example handles three classes and you need a perceptron with three outputs. This is equivalent to three single-output perceptrons in parallel.

The third example considers classification in a three-dimensional space. The main difference from the two-dimensional classification is that you cannot properly plot the data.

Notice that if you reevaluate the examples you will not receive exactly the same results due to the random initialization used in training perceptron networks. See Section 4.1.1, InitializePerceptron.