Perceptrons are simple single-layer binary classifiers, which divide the input space with a linear decision boundary.
Perceptrons can learn to solve a narrow range of classification problems. They were one of the first neural networks to reliably solve a given class of problem, and their advantage is a simple learning rule.
perceptron(hardlimitTF,perceptronLF) takes these arguments,
Hard limit transfer function (default = 'hardlim')
Perceptron learning rule (default = 'learnp')
and returns a perceptron.
Note Neural Network Toolbox™ supports perceptrons for historical interest. For better results, you should instead use patternnet, which can solve nonlinearly separable problems. Sometimes the term "perceptrons" refers to feed-forward pattern recognition networks; but the original perceptron, described here, can solve only simple problems.
Use a perceptron to solve a simple classification logical-OR problem.
x = [0 0 1 1; 0 1 0 1]; t = [0 1 1 1]; net = perceptron; net = train(net,x,t); view(net) y = net(x);