The path to dataset can be provided using the dataset
parameter and mode
parameter can be used to specify the mode in which to execute perceptron. There are two modes available: erm
for Empirical Risk Minimization and cv
for 10 fold Cross Validation. For example:
python perceptron.py --dataset 'path/to/dataset' --mode erm