As mentioned in Perceptron.ipynb, the gradient descent algorithm comes with a learning rate. However there is no learning rate in the train function.