How to make an Averaged Perceptron using Shogun?

Recipe Objective

This recipe explains how to make an Averaged Perceptron using Shogun.
For more related projects-
Project 1
Project 2

Learn to Implement Customer Churn Prediction Using Machine Learning in Python

Averaged Perceptron

The averaged Perceptron uses the averaged weight and bias. It is an online binary classifier which is an extension of the standard Perceptron algorithm. Given a vector, a predicted class is given by:
f(weight. a + bias)
Here, weight is the average weight vector, bias is the average bias and f() is a step function where step function is equal to 1 for 'a' greater than 0, equal to -1 for 'a' less than 0, and equal to zero for 'a' equal to zero.

x_train = RealFeatures(feats_train)
x_test = RealFeatures(feats_test)
y_train = BinaryLabels(labels_train)
y_test = BinaryLabels(labels_test)

lr = 1.0
maxx = 1000
perceptron = AveragedPerceptron(x_train, y_train)
perceptron.set_learn_rate(y_rate)
perceptron.set_max_iter(maxx)

perceptron.train()
perceptron.set_features(x_test)
predict = perceptron.apply()

eval = AccuracyMeasure()
accuracy = eval.evaluate(predict, y_test)