An implementation of a Perceptron using early stopping to prevent over-fitting. Used the step function as the activation function.

• The program takes arguments in the form: [train-set][tune-set][test-set].

#Input file format:

1. The files are expected to have the following format:

• The first line, an integer for the number of features per example.
• Next, list each feature on its own line with the its possible values, ex. gender – male female.
• Next, the two label features are listed, each on its own line.
• Next, an integer for the number of examples in the file, on its own line.
• Lastly, the examples, listed as the example each on its own line, ex.[example name][label][feature values]. The feature values should have the same order as they were given to the program.

2. The program will ignore empty lines and lines that start with “//”.
3. The program takes binary features and labels. If the set has numerical values, this is easy to work around, just have a threshold for each value.
4. Change the step and patience values according to the desired error value.