Minimising the perceptron criterion function

For creating an example to demonstrate how to train an artificial neuron with a simple training algorithm I created a spreadsheet that trains the neuron given a truth table, initial weights and a learning rate.

It therefore uses the perceptron criterion function to train a single neuron, by adding the values of misclassified inputs to the weights in batch mode.

You can play around with different learning rates, initial weights and different functions. Each step is shown if it's adaption of the weights.

See attached file (OASIS spreadsheet).

AttachmentSize
Perceptron_criterion_function.ods17.19 KB