

In [1]:

    
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

from adaline import AdalineGD, AdalineSGD
from graphhelpers import plot_decision_regions

%matplotlib inline

Data Preparation



In [2]:

    
col_names = ['sepal length (cm)', 'sepal width (cm)', 'petal length (cm)', 'petal width (cm)', 'class']
df = pd.read_csv('../data/iris/iris.data', header=None, names=col_names)
y = df.iloc[:100,4].values  # Grab only the first 2 classifications (first 100 rows)
y = np.where(y == 'Iris-setosa', -1, 1)
X = df.iloc[:100,[0,2]].values  # Grab only the 1st and 3rd column

Adaline Learning Rate & Errors



In [3]:

    
_, (left_plot, right_plot) = plt.subplots(nrows=1, ncols=2, figsize=(8, 4))

ada1 = AdalineGD(iterations=10, learning_rate=.01).fit(X, y)
left_plot.plot(range(1, len(ada1.cost) + 1), np.log10(ada1.cost), marker='o')
left_plot.set_xlabel('Epochs')
left_plot.set_ylabel('log(Sum of Squared Errors)')
left_plot.set_title('Adaline - 0.01 Learning Rate')

ada2 = AdalineGD(iterations=10, learning_rate=.0001).fit(X, y)
right_plot.plot(range(1, len(ada2.cost) + 1), ada2.cost, marker='o')
right_plot.set_xlabel('Epochs')
right_plot.set_ylabel('Sum of Sqaured Errors')
right_plot.set_title('Adaline - 0.0001 Learning Rate')
plt.show()

Standardization (Type of Feature Scaling)

This helps with gradient decent (along with other algorithms)



In [4]:

    
X_std = np.copy(X)
for c in xrange(X.shape[1]):
    X_std[:, c] = (X[:, c] - X[:, c].mean()) / X[:, c].std()

Adaline After Standardization



In [5]:

    
ada = AdalineGD(iterations=15, learning_rate=.01)
ada.fit(X_std, y)

plot_decision_regions(X_std, y, classifier=ada)
plt.title('Adaline - Gradient Descent')
plt.xlabel('sepal length (standardized)')
plt.ylabel('petal length (standardized)')
plt.legend(loc='upper left')
plt.show()

plt.plot(range(1, len(ada.cost) + 1), ada.cost, marker='o')
plt.xlabel('Epochs')
plt.ylabel('Sum of Squared Erorrs')









    












    Out[5]:





<matplotlib.text.Text at 0x11179e950>

Adaline - Stochastic Gradient Descent



In [6]:

    
ada_sgd = AdalineSGD(iterations=15, learning_rate=.01, random_state=1)
ada_sgd.fit(X_std, y)









    Out[6]:





<adaline.AdalineSGD at 0x111658050>



In [7]:

    
plot_decision_regions(X_std, y, classifier=ada_sgd)
plt.title('Adaline - Stochastic Gradient Descent')
plt.xlabel('sepal length (standardized)')
plt.ylabel('petal length (standardized)')
plt.legend(loc='upper left')
plt.show()

plt.plot(range(1, len(ada_sgd.cost) + 1), ada_sgd.cost, marker='o')
plt.xlabel('Epochs')
plt.ylabel('Average Cost')
plt.show()

Online Learning



In [8]:

    
ada_sgd.partial_fit(X_std[0, :], y[0])  # only 1 data point









    Out[8]:





<adaline.AdalineSGD at 0x111658050>



In [ ]: