In [1]:

    

import tensorflow as tf
import numpy as np
rng = np.random

import matplotlib.pyplot as plt
learning_rate = 0.0001
training_epochs = 1000
display_step = 50


    

In [2]:

    

with tf.name_scope("Creation_of_array"):
    x_array=np.asarray([2.0,9.4,3.32,0.88,-2.23,1.11,0.57,-2.25,-3.31,6.45])
    y_array=np.asarray([1.22,0.34,-0.08,2.25,4.41,3.09,-6.66,-9.77,0.001,2.25])
    x = tf.constant(x_array,dtype = tf.float32,name = "x_array")
    y = tf.constant(y_array,dtype = tf.float32, name= "y_array")
with tf.name_scope("Calculating_y_mean"):
    mean_y = tf.reduce_mean(y, name = "mean_y")
    with tf.Session() as sess:
        result_y = sess.run(mean_y)
        print(result_y)


    

    




-0.2949

In [3]:

    

with tf.name_scope("Calculating_x_mean_and_x_variance"):
    mean_x, variance = tf.nn.moments(x, [0], name = "mean_x_and_variance_x")
    with tf.Session() as sess:
        m, v = sess.run([mean_x, variance])
        print(m)
        print(v)


    

    




1.594
14.2899

In [4]:

    

with tf.name_scope("Calculating_covariance"):
    def tensorflow_covariance(x_array,y_array,x_mean,y_mean):
        cov = 0.0
        for i in range(0,10):
            x_val = tf.subtract(x_array[i],x_mean, name="Finding_difference_of_xval_and_mean")
            y_val = tf.subtract(y_array[i],y_mean, name="Finding_difference_of_yval_and_mean")
            total_val = tf.multiply(x_val,y_val, name="Multiplying_found_values")
            cov = tf.add(cov,total_val, name="Recursive_addition")
        return cov/10.0
    with tf.Session() as sess:
        covar = sess.run(tensorflow_covariance(x,y,m,result_y))
        print(covar)


    

    




3.83422

In [5]:

    

with tf.name_scope("Calculating_slope_m_and_c"):
    slope = tf.div(covar,v,name="Finding_slope")
    intm = tf.multiply(slope,m,name = "Intermediate_step")
    c_intm = tf.subtract(result_y,intm,name = "Finding_c")

    with tf.Session() as sess:
        m_slope = sess.run(slope)
        c = sess.run(c_intm)
        print(m_slope)
        print(c)


    

    




0.268316
-0.722596

In [6]:

    

with tf.name_scope("Plotting"):
    n_samples = x_array.shape[0]
    X = tf.placeholder("float")
    Y = tf.placeholder("float")

    # Set model weights
    W = tf.Variable(rng.randn(), name="weight")
    b = tf.Variable(rng.randn(), name="bias")

    # Construct a linear model
    pred = tf.add(tf.multiply(X, W), b)


    # Mean squared error
    cost = tf.reduce_sum(tf.pow(pred-Y, 2))/(2*n_samples)
    # Gradient descent
    optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

    # Initializing the variables
    init = tf.global_variables_initializer()

    # Launch the graph
    with tf.Session() as sess:
        sess.run(init)

        # Fit all training data
        for epoch in range(training_epochs):
            for (p, r) in zip(x_array, y_array):
                sess.run(optimizer, feed_dict={X: p, Y: r})

            # Display logs per epoch step
            if (epoch+1) % display_step == 0:
                c = sess.run(cost, feed_dict={X: x_array, Y:y_array})
                print("Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(c), \
                    "W=", sess.run(W), "b=", sess.run(b))

        print("Optimization Finished!")
        training_cost = sess.run(cost, feed_dict={X: x_array, Y: y_array})
        print("Training cost=", training_cost, "W=", sess.run(W), "b=", sess.run(b), '\n')

        # Graphic display
        plt.plot(x_array, y_array, 'ro', label='Original data')
        plt.plot(x_array, sess.run(W) * x_array + sess.run(b), label='Fitted line')
        plt.legend()
        plt.show()


    

    




Epoch: 0050 cost= 14.431994438 W= 1.08169 b= -0.446887
Epoch: 0100 cost= 13.502418518 W= 1.01395 b= -0.454459
Epoch: 0150 cost= 12.718105316 W= 0.951728 b= -0.461476
Epoch: 0200 cost= 12.056349754 W= 0.894586 b= -0.467983
Epoch: 0250 cost= 11.497995377 W= 0.842104 b= -0.474023
Epoch: 0300 cost= 11.026884079 W= 0.793903 b= -0.479632
Epoch: 0350 cost= 10.629384995 W= 0.749636 b= -0.484846
Epoch: 0400 cost= 10.293990135 W= 0.708981 b= -0.489696
Epoch: 0450 cost= 10.010996819 W= 0.671643 b= -0.494212
Epoch: 0500 cost= 9.772211075 W= 0.637354 b= -0.49842
Epoch: 0550 cost= 9.570723534 W= 0.605863 b= -0.502347
Epoch: 0600 cost= 9.400713921 W= 0.576945 b= -0.506013
Epoch: 0650 cost= 9.257253647 W= 0.550388 b= -0.509442
Epoch: 0700 cost= 9.136195183 W= 0.526 b= -0.51265
Epoch: 0750 cost= 9.034040451 W= 0.503605 b= -0.515656
Epoch: 0800 cost= 8.947832108 W= 0.48304 b= -0.518477
Epoch: 0850 cost= 8.875081062 W= 0.464157 b= -0.521127
Epoch: 0900 cost= 8.813679695 W= 0.446817 b= -0.523619
Epoch: 0950 cost= 8.761857986 W= 0.430896 b= -0.525967
Epoch: 1000 cost= 8.718118668 W= 0.416278 b= -0.528183
Optimization Finished!
Training cost= 8.71812 W= 0.416278 b= -0.528183 







    

In [ ]:

    




    

In [7]:

    

with tf.name_scope("Finding_root_mean_square_error"):
    rms = tf.sqrt(tf.reduce_mean(tf.squared_difference(x_array, y_array,name = "Finding_squared_difference"),name="Finding_mean"),name = "Finding_square_root")
    with tf.Session() as sess:
        rmse=sess.run(rms)
        print(rmse)


    

    




5.30626913189

In [8]:

    

with tf.name_scope("Finding_theta_1"): 
    y_var = tf.subtract(y,result_y,name = "Subtract_y_array_with_y_mean")
    x_var = tf.subtract(x,m,name = "Subtract_x_array_with_x_mean")
    mult = tf.multiply(x_var,y_var,name = "Multiply_calculated_arrays")
    sumn = tf.reduce_sum(mult,name = "Find_sum_of_x_i_minus_mean_x_and_y_i_minus_mean_y")
    x_var2 = tf.multiply(x_var,x_var,name = "Squaring_found_arrray_values")
    sumd = tf.reduce_sum(x_var2,name = "Find_sum_of_array_of_x_i_minus_mean_x")
    val = sumn/sumd

    with tf.Session() as sess:
        res = sess.run(val)
        print(res)


    

    




0.268316

In [9]:

    

with tf.name_scope("Finding_theta_0"):    
    temp = tf.multiply(res,m,name = "Multiply_res_with_slope")
    theta = tf.subtract(result_y,temp,name="Sub_obtained_res_with_mean_y")
    with tf.Session() as sess:
        theta0 = sess.run(theta)
        print(theta0)


    

    




-0.722596

In [ ]: