Learning from Data: Perceptron Exercise

perceptron.py

import numpy as np

# Find f(x), where f is the line through p1 and p2
def f(p1, p2, x):
    return (p2[1]-p1[1])/(p2[0]-p1[0])*(x-p1[0])+p1[1]
  
# Get random points p1 and p2
p1, p2 = np.random.rand(2, 2)
line = lambda x: f(p1, p2, x)

def label(p):
    return np.sign(p[1] - line(p[0]))
  
# Get number of training sample points to define
n_train = int(input("Input the size of the training set\n"))

def run_sample():
    w = np.array([0.0, 0.0, 0.0])
    train_sample = np.random.rand(n_train, 2)
    n_label = np.array([label(p) for p in train_sample])
    misclassified = list(range(n_train))
    num_iters = 0
    while misclassified:
        num_iters += 1
        np.random.shuffle(misclassified)
        i = misclassified[0]
        w[0] += n_label[i]*train_sample[i][0]
        w[1] += n_label[i]*train_sample[i][1]
        w[2] += n_label[i]
        temp = []
        for i in range(n_train):
            if n_label[i] != np.sign(w[0]*train_sample[i][0]+w[1]*train_sample[i][1]+w[2]):
                temp.append(i)
        misclassified = temp
    return num_iters, w
  
iters = []
num_wrong = 0
for _ in range(1000):
    n, w = run_sample()
    iters.append(n)
    p = np.random.rand(2)
    if label(p) != np.sign(w[0]*p[0]+w[1]*p[1]+w[2]):
        num_wrong += 1
print(np.average(iters))
print(num_wrong)