In [1]:

    

import pandas as pd


    

In [2]:

    

pd.__version__


    

    
Out[2]:





'0.14.1'

In [3]:

    

import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf


    

In [4]:

    

tf.logging.set_verbosity(tf.logging.ERROR)
%matplotlib inline
import warnings; warnings.simplefilter('ignore')


    

In [5]:

    

my_feature = pd.Series([1, 3, 5, 7, 9])
my_targets = pd.Series([12, 13, 14, 15, 16])


    

In [6]:

    

feature_columns = tf.contrib.learn.infer_real_valued_columns_from_input(my_feature)


    

In [7]:

    

linear_regressor = tf.contrib.learn.LinearRegressor(
    feature_columns=feature_columns,
    optimizer=tf.train.GradientDescentOptimizer(learning_rate=0.00001),
    enable_centered_bias=True,
    gradient_clip_norm=5.0
)


    

In [8]:

    

print "Training model..."
linear_regressor.fit(
    my_feature,
    my_targets,
    steps=1000,
    batch_size=15
)
print "Model training finished."


    

    




Training model...
Model training finished.

In [9]:

    

predictions = list(linear_regressor.predict(my_feature, as_iterable=True))

print predictions


    

    




[5.6934004, 5.7917185, 5.8900361, 5.9883542, 6.0866718]

In [10]:

    

plt.ylabel('targets')
plt.scatter(my_feature, my_targets, c='blue')
plt.scatter(my_feature, predictions, c='red')
plt.show()


    

In [ ]: