In [25]:

    

%matplotlib inline


    

In [26]:

    

from code import perceptron


    

In [27]:

    

data = pd.read_csv('data/simple.iris.csv')
data.head(5)


    

    
Out[27]:






  

    

      

      
f1
      
f2
      
y
    
  
  

    

      
0
      
5.1
      
1.4
      
-1.0
    
    

      
1
      
4.9
      
1.4
      
-1.0
    
    

      
2
      
4.7
      
1.3
      
-1.0
    
    

      
3
      
4.6
      
1.5
      
-1.0
    
    

      
4
      
5.0
      
1.4
      
-1.0
    
  

In [28]:

    

X = data.values[:, :2]
y = data.values[:, 2]
y = y.reshape(len(y), 1)[:, 0]
weights, errors = perceptron.fit(X, y)


    

In [29]:

    

weights


    

    
Out[29]:





array([-0.04 , -0.068,  0.182])

In [30]:

    

errors


    

    
Out[30]:





[2, 2, 3, 2, 1, 0, 0, 0, 0, 0]

In [31]:

    

import matplotlib.pyplot as plt
import seaborn as sns


    

In [32]:

    

xs = range(1, len(errors) + 1)
plt.plot(xs, errors, marker='o')
plt.ylim(min(errors) - 0.5, max(errors) + 0.5)
plt.xlim(min(xs) - 0.5, max(xs) + 0.5)
plt.xlabel('training epoch')
plt.ylabel('classification errors')


    

    
Out[32]:





<matplotlib.text.Text at 0x115947b50>






    

In [33]:

    

from matplotlib.colors import ListedColormap
def plot_decision_regions(X, y, classifier, resolution=0.02):
    markers = ('s', 'x', 'o', '^', 'v')
    cmap = ListedColormap(sns.color_palette()[:len(np.unique(y))])
    x1_min, x1_max = X[:, 0].min() - 1, X[:, 0].max() + 1
    x2_min, x2_max = X[:, 1].min() - 1, X[:, 1].max() + 1
    xx1, xx2 = np.meshgrid(np.arange(x1_min, x1_max, resolution),
                         np.arange(x2_min, x2_max, resolution))
    Z = classifier.predict(np.array([xx1.ravel(), xx2.ravel()]).T, weights)
    Z = Z.reshape(xx1.shape)
    plt.contourf(xx1, xx2, Z, alpha=0.4, cmap=cmap)
    plt.xlim(xx1.min(), xx1.max())
    plt.ylim(xx2.min(), xx2.max())
    # plot class samples
    for idx, cl in enumerate(np.unique(y)):
        plt.scatter(x=X[y == cl, 0], y=X[y == cl, 1],
                    alpha=0.8, c=cmap(idx),
                    marker=markers[idx], label=cl)


    

In [35]:

    

plot_decision_regions(X, y, classifier=perceptron)
plt.legend(loc='upper left')


    

    
Out[35]:





<matplotlib.legend.Legend at 0x115c0be50>