In [1]:

    

from __future__ import division
import tensorflow as tf
import numpy as np
import tarfile
import os
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import time
get_ipython().magic('matplotlib inline')



def csv_to_numpy_array(filePath, delimiter):
    return np.genfromtxt(filePath, delimiter=delimiter, dtype=None)

def import_data():
    if "data" not in os.listdir(os.getcwd()):
        # Untar directory of data if we haven't already
        tarObject = tarfile.open("data.tar.gz")
        tarObject.extractall()
        tarObject.close()
        print("Extracted tar to current directory")
    else:
        # we've already extracted the files
        pass

    print("loading training data")
    trainX = csv_to_numpy_array("data/trainX.csv", delimiter="\t")
    trainY = csv_to_numpy_array("data/trainY.csv", delimiter="\t")
    print("loading test data")
    testX = csv_to_numpy_array("data/testX.csv", delimiter="\t")
    testY = csv_to_numpy_array("data/testY.csv", delimiter="\t")
    return trainX,trainY,testX,testY

trainX,trainY,testX,testY = import_data()


    

    




loading training data
loading test data

In [2]:

    

numFeatures = trainX.shape[1]
numLabels = trainY.shape[1]

numEpochs = 20000
learningRate = 0.5
print ("Num Features",numFeatures)
print ("Labels:",numLabels)


    

    




Num Features 2955
Labels: 2

In [3]:

    

X = tf.placeholder(tf.float32, [None, numFeatures])
yGold = tf.placeholder(tf.float32, [None, numLabels])

# Hidden Layer 1 Parameters
weights1 = tf.Variable(tf.random_normal([numFeatures,numLabels*3],
                                       mean=0,
                                       stddev=(np.sqrt(6/numFeatures+
                                                         numLabels+1)),
                                       name="weights"))

bias1 = tf.Variable(tf.random_normal([1,numLabels*3],
                                    mean=0,
                                    stddev=(np.sqrt(6/numFeatures+numLabels+1)),
                                    name="bias"))
# Hidden Layer 2 parameters
weights2 = tf.Variable(tf.random_normal([numLabels*3,numLabels],
                                       mean=0,
                                       stddev=(np.sqrt(6/numFeatures+
                                                         numLabels+1)),
                                       name="weights"))

bias2 = tf.Variable(tf.random_normal([1,numLabels],
                                    mean=0,
                                    stddev=(np.sqrt(6/numFeatures+numLabels+1)),
                                    name="bias"))


    

In [5]:

    

init_OP = tf.initialize_all_variables()

# Hidden Layer 1: Obtaining weighted matrix
apply_weights_OP = tf.matmul(X, weights1, name="apply_weights")
add_bias_OP = tf.add(apply_weights_OP, bias1, name="add_bias") 

# Activation Funciton
activation_OP1 = tf.nn.sigmoid(add_bias_OP, name="activation")

# Hidden Layer 2
activation_OP = tf.nn.softmax(tf.add(tf.matmul(activation_OP1,weights2), bias2))


# Cross Entropy eror
#cost_OP = tf.nn.l2_loss(activation_OP-yGold, name="squared_error_cost")
cost_OP = tf.reduce_mean(-tf.reduce_sum(yGold * tf.log(activation_OP), reduction_indices=[1]))

# Gradient Descent
training_OP = tf.train.GradientDescentOptimizer(learningRate).minimize(cost_OP)

epoch_values=[]
accuracy_values=[]
cost_values=[]
# Turn on interactive plotting
plt.ion()
# Create the main, super plot
fig = plt.figure()
# Create two subplots on their own axes and give titles
ax1 = plt.subplot("211")
ax1.set_title("TRAINING ACCURACY", fontsize=18)
ax2 = plt.subplot("212")
ax2.set_title("TRAINING COST", fontsize=18)
plt.tight_layout()


# Initialize reporting variables
cost = 0
diff = 1
errors =[]
with tf.Session() as sess:
    sess.run(init_OP)
    correct_predictions_OP = tf.equal(tf.argmax(activation_OP,1),tf.argmax(yGold,1))
    accuracy_OP = tf.reduce_mean(tf.cast(correct_predictions_OP, "float"))
        

# Training
    for i in range(numEpochs):
        if i > 1 and diff < .0001:
            print("change in cost %g; convergence."%diff)
            break
        else:
        # Run training step
            step = sess.run(training_OP, feed_dict={X: trainX, yGold: trainY})
        # Report occasional stats
            if i % 100 == 0:
            # Add epoch to epoch_values
                epoch_values.append(i)
            # Generate accuracy stats on test data
                train_accuracy, newCost = sess.run(
                [accuracy_OP, cost_OP], 
                feed_dict={X: trainX, yGold: trainY}
                )
                errors.append(1 - train_accuracy)
            # Add accuracy to live graphing variable
                accuracy_values.append(train_accuracy)
            # Add cost to live graphing variable
                cost_values.append(newCost)
           
            # Re-assign values for variables
                diff = abs(newCost - cost)
                cost = newCost

            #generate print statements
                print("step %d, training accuracy %g"%(i, train_accuracy))
                #print("step %d, cost %g"%(i, newCost))
                print("step %d, change in cost %g"%(i, diff))

            # Plot progress to our two subplots
                accuracyLine, = ax1.plot(epoch_values, accuracy_values)
                costLine, = ax2.plot(epoch_values, cost_values)
                fig.canvas.draw()
                time.sleep(1)


    print("final accuracy on test set: %s" %str(sess.run(accuracy_OP, 
                                                     feed_dict={X: testX, 
                                                                yGold: testY})))
    print("Error: ", errors[-1])
    
    plt.title('Error Plot')
    plt.plot([np.mean(errors[i-50:i]) for i in range(len(errors))])
    plt.show()

# Create Saver
saver = tf.train.Saver()
# Save variables to .ckpt file
# saver.save(sess, "trained_variables.ckpt")


    

    




step 0, training accuracy 0.466946
step 0, change in cost 1.59345
step 100, training accuracy 0.585519
step 100, change in cost 0.919326
step 200, training accuracy 0.66212
step 200, change in cost 0.0201554
step 300, training accuracy 0.714586
step 300, change in cost 0.0194233
step 400, training accuracy 0.746065
step 400, change in cost 0.0210141
step 500, training accuracy 0.767051
step 500, change in cost 0.0237725
step 600, training accuracy 0.789087
step 600, change in cost 0.0270252
step 700, training accuracy 0.820567
step 700, change in cost 0.0301889
step 800, training accuracy 0.837356
step 800, change in cost 0.0326526
step 900, training accuracy 0.869885
step 900, change in cost 0.0338979
step 1000, training accuracy 0.889822
step 1000, change in cost 0.0337308
step 1100, training accuracy 0.902413
step 1100, change in cost 0.0323577
step 1200, training accuracy 0.70829
step 1200, change in cost 0.0785539
step 1300, training accuracy 0.770199
step 1300, change in cost 0.0430034
step 1400, training accuracy 0.78384
step 1400, change in cost 0.0188711
step 1500, training accuracy 0.797482
step 1500, change in cost 0.0184191
step 1600, training accuracy 0.820567
step 1600, change in cost 0.01845
step 1700, training accuracy 0.835257
step 1700, change in cost 0.0186076
step 1800, training accuracy 0.850997
step 1800, change in cost 0.0189115
step 1900, training accuracy 0.873033
step 1900, change in cost 0.0194111
step 2000, training accuracy 0.889822
step 2000, change in cost 0.0201961
step 2100, training accuracy 0.900315
step 2100, change in cost 0.021417
step 2200, training accuracy 0.916055
step 2200, change in cost 0.023347
step 2300, training accuracy 0.925498
step 2300, change in cost 0.0262975
step 2400, training accuracy 0.945435
step 2400, change in cost 0.0284637
step 2500, training accuracy 0.954879
step 2500, change in cost 0.0222528
step 2600, training accuracy 0.959077
step 2600, change in cost 0.0126056
step 2700, training accuracy 0.965373
step 2700, change in cost 0.00834189
step 2800, training accuracy 0.96852
step 2800, change in cost 0.0069409
step 2900, training accuracy 0.96852
step 2900, change in cost 0.00630465
step 3000, training accuracy 0.96852
step 3000, change in cost 0.00581549
step 3100, training accuracy 0.96852
step 3100, change in cost 0.00539014
step 3200, training accuracy 0.96852
step 3200, change in cost 0.00501567
step 3300, training accuracy 0.970619
step 3300, change in cost 0.00468406
step 3400, training accuracy 0.971668
step 3400, change in cost 0.00438841
step 3500, training accuracy 0.971668
step 3500, change in cost 0.00412308
step 3600, training accuracy 0.972718
step 3600, change in cost 0.00388367
step 3700, training accuracy 0.973767
step 3700, change in cost 0.00366608
step 3800, training accuracy 0.973767
step 3800, change in cost 0.00346721
step 3900, training accuracy 0.974816
step 3900, change in cost 0.00328428
step 4000, training accuracy 0.974816
step 4000, change in cost 0.00311501
step 4100, training accuracy 0.974816
step 4100, change in cost 0.00295757
step 4200, training accuracy 0.975866
step 4200, change in cost 0.00281054
step 4300, training accuracy 0.975866
step 4300, change in cost 0.00267281
step 4400, training accuracy 0.976915
step 4400, change in cost 0.00254346
step 4500, training accuracy 0.977964
step 4500, change in cost 0.00242201
step 4600, training accuracy 0.977964
step 4600, change in cost 0.00230791
step 4700, training accuracy 0.979014
step 4700, change in cost 0.00220085
step 4800, training accuracy 0.980063
step 4800, change in cost 0.00210042
step 4900, training accuracy 0.980063
step 4900, change in cost 0.00200618
step 5000, training accuracy 0.981112
step 5000, change in cost 0.00191795
step 5100, training accuracy 0.981112
step 5100, change in cost 0.00183512
step 5200, training accuracy 0.98426
step 5200, change in cost 0.00175747
step 5300, training accuracy 0.98426
step 5300, change in cost 0.00168453
step 5400, training accuracy 0.98426
step 5400, change in cost 0.00161598
step 5500, training accuracy 0.98426
step 5500, change in cost 0.00155144
step 5600, training accuracy 0.986359
step 5600, change in cost 0.00149065
step 5700, training accuracy 0.987408
step 5700, change in cost 0.00143329
step 5800, training accuracy 0.987408
step 5800, change in cost 0.00137908
step 5900, training accuracy 0.987408
step 5900, change in cost 0.00132784
step 6000, training accuracy 0.987408
step 6000, change in cost 0.0012793
step 6100, training accuracy 0.987408
step 6100, change in cost 0.00123327
step 6200, training accuracy 0.988458
step 6200, change in cost 0.00118964
step 6300, training accuracy 0.989507
step 6300, change in cost 0.00114815
step 6400, training accuracy 0.989507
step 6400, change in cost 0.00110871
step 6500, training accuracy 0.989507
step 6500, change in cost 0.00107121
step 6600, training accuracy 0.989507
step 6600, change in cost 0.00103546
step 6700, training accuracy 0.990556
step 6700, change in cost 0.0010014
step 6800, training accuracy 0.990556
step 6800, change in cost 0.000968948
step 6900, training accuracy 0.990556
step 6900, change in cost 0.000937942
step 7000, training accuracy 0.990556
step 7000, change in cost 0.000908341
step 7100, training accuracy 0.990556
step 7100, change in cost 0.000880066
step 7200, training accuracy 0.991605
step 7200, change in cost 0.000853028
step 7300, training accuracy 0.991605
step 7300, change in cost 0.000827149
step 7400, training accuracy 0.991605
step 7400, change in cost 0.000802387
step 7500, training accuracy 0.991605
step 7500, change in cost 0.000778683
step 7600, training accuracy 0.991605
step 7600, change in cost 0.00075594
step 7700, training accuracy 0.992655
step 7700, change in cost 0.000734158
step 7800, training accuracy 0.992655
step 7800, change in cost 0.000713259
step 7900, training accuracy 0.992655
step 7900, change in cost 0.000693206
step 8000, training accuracy 0.992655
step 8000, change in cost 0.000673965
step 8100, training accuracy 0.992655
step 8100, change in cost 0.000655472
step 8200, training accuracy 0.992655
step 8200, change in cost 0.000637718
step 8300, training accuracy 0.993704
step 8300, change in cost 0.000620648
step 8400, training accuracy 0.993704
step 8400, change in cost 0.000604223
step 8500, training accuracy 0.993704
step 8500, change in cost 0.000588439
step 8600, training accuracy 0.993704
step 8600, change in cost 0.000573225
step 8700, training accuracy 0.994753
step 8700, change in cost 0.000558611
step 8800, training accuracy 0.994753
step 8800, change in cost 0.000544488
step 8900, training accuracy 0.994753
step 8900, change in cost 0.000530928
step 9000, training accuracy 0.994753
step 9000, change in cost 0.000517838
step 9100, training accuracy 0.994753
step 9100, change in cost 0.000505231
step 9200, training accuracy 0.994753
step 9200, change in cost 0.000493057
step 9300, training accuracy 0.994753
step 9300, change in cost 0.000481304
step 9400, training accuracy 0.994753
step 9400, change in cost 0.000469986
step 9500, training accuracy 0.994753
step 9500, change in cost 0.000459034
step 9600, training accuracy 0.994753
step 9600, change in cost 0.000448484
step 9700, training accuracy 0.994753
step 9700, change in cost 0.000438269
step 9800, training accuracy 0.994753
step 9800, change in cost 0.000428423
step 9900, training accuracy 0.994753
step 9900, change in cost 0.00041889
step 10000, training accuracy 0.995803
step 10000, change in cost 0.000409674
step 10100, training accuracy 0.995803
step 10100, change in cost 0.000400767
step 10200, training accuracy 0.995803
step 10200, change in cost 0.000392139
step 10300, training accuracy 0.996852
step 10300, change in cost 0.000383794
step 10400, training accuracy 0.996852
step 10400, change in cost 0.000375733
step 10500, training accuracy 0.996852
step 10500, change in cost 0.000367897
step 10600, training accuracy 0.996852
step 10600, change in cost 0.000360299
step 10700, training accuracy 0.996852
step 10700, change in cost 0.000352977
step 10800, training accuracy 0.996852
step 10800, change in cost 0.000345876
step 10900, training accuracy 0.996852
step 10900, change in cost 0.00033897
step 11000, training accuracy 0.996852
step 11000, change in cost 0.000332294
step 11100, training accuracy 0.996852
step 11100, change in cost 0.000325803
step 11200, training accuracy 0.996852
step 11200, change in cost 0.000319531
step 11300, training accuracy 0.996852
step 11300, change in cost 0.000313418
step 11400, training accuracy 0.996852
step 11400, change in cost 0.000307487
step 11500, training accuracy 0.996852
step 11500, change in cost 0.000301734
step 11600, training accuracy 0.996852
step 11600, change in cost 0.000296157
step 11700, training accuracy 0.996852
step 11700, change in cost 0.000290737
step 11800, training accuracy 0.996852
step 11800, change in cost 0.000285443
step 11900, training accuracy 0.996852
step 11900, change in cost 0.00028035
step 12000, training accuracy 0.996852
step 12000, change in cost 0.000275362
step 12100, training accuracy 0.996852
step 12100, change in cost 0.000270508
step 12200, training accuracy 0.996852
step 12200, change in cost 0.000265829
step 12300, training accuracy 0.996852
step 12300, change in cost 0.000261215
step 12400, training accuracy 0.996852
step 12400, change in cost 0.000256758
step 12500, training accuracy 0.996852
step 12500, change in cost 0.000252441
step 12600, training accuracy 0.996852
step 12600, change in cost 0.000248183
step 12700, training accuracy 0.996852
step 12700, change in cost 0.000244085
step 12800, training accuracy 0.996852
step 12800, change in cost 0.000240084
step 12900, training accuracy 0.996852
step 12900, change in cost 0.000236176
step 13000, training accuracy 0.996852
step 13000, change in cost 0.000232371
step 13100, training accuracy 0.996852
step 13100, change in cost 0.000228681
step 13200, training accuracy 0.996852
step 13200, change in cost 0.000225043
step 13300, training accuracy 0.997901
step 13300, change in cost 0.000221524
step 13400, training accuracy 0.997901
step 13400, change in cost 0.000218088
step 13500, training accuracy 0.997901
step 13500, change in cost 0.000214737
step 13600, training accuracy 0.997901
step 13600, change in cost 0.000211427
step 13700, training accuracy 0.997901
step 13700, change in cost 0.000208257
step 13800, training accuracy 0.997901
step 13800, change in cost 0.000205141
step 13900, training accuracy 0.997901
step 13900, change in cost 0.000202103
step 14000, training accuracy 0.997901
step 14000, change in cost 0.000199098
step 14100, training accuracy 0.997901
step 14100, change in cost 0.000196179
step 14200, training accuracy 0.997901
step 14200, change in cost 0.000193333
step 14300, training accuracy 0.997901
step 14300, change in cost 0.000190575
step 14400, training accuracy 0.997901
step 14400, change in cost 0.00018785
step 14500, training accuracy 0.997901
step 14500, change in cost 0.000185175
step 14600, training accuracy 0.997901
step 14600, change in cost 0.000182571
step 14700, training accuracy 0.997901
step 14700, change in cost 0.000180034
step 14800, training accuracy 0.997901
step 14800, change in cost 0.000177545
step 14900, training accuracy 0.997901
step 14900, change in cost 0.000175091
step 15000, training accuracy 0.997901
step 15000, change in cost 0.000172714
step 15100, training accuracy 0.997901
step 15100, change in cost 0.000170367
step 15200, training accuracy 0.997901
step 15200, change in cost 0.000168063
step 15300, training accuracy 0.997901
step 15300, change in cost 0.000165815
step 15400, training accuracy 0.997901
step 15400, change in cost 0.000163618
step 15500, training accuracy 0.997901
step 15500, change in cost 0.000161458
step 15600, training accuracy 0.997901
step 15600, change in cost 0.000159327
step 15700, training accuracy 0.997901
step 15700, change in cost 0.000157259
step 15800, training accuracy 0.997901
step 15800, change in cost 0.000155214
step 15900, training accuracy 0.997901
step 15900, change in cost 0.000153221
step 16000, training accuracy 0.997901
step 16000, change in cost 0.000151252
step 16100, training accuracy 0.997901
step 16100, change in cost 0.000149304
step 16200, training accuracy 0.997901
step 16200, change in cost 0.000147415
step 16300, training accuracy 0.997901
step 16300, change in cost 0.00014554
step 16400, training accuracy 0.997901
step 16400, change in cost 0.000143725
step 16500, training accuracy 0.997901
step 16500, change in cost 0.000141941
step 16600, training accuracy 0.997901
step 16600, change in cost 0.00014017
step 16700, training accuracy 0.997901
step 16700, change in cost 0.0001384
step 16800, training accuracy 0.997901
step 16800, change in cost 0.000136705
step 16900, training accuracy 0.997901
step 16900, change in cost 0.000135038
step 17000, training accuracy 0.997901
step 17000, change in cost 0.000133406
step 17100, training accuracy 0.997901
step 17100, change in cost 0.000131737
step 17200, training accuracy 0.997901
step 17200, change in cost 0.000130141
step 17300, training accuracy 0.997901
step 17300, change in cost 0.000128573
step 17400, training accuracy 0.997901
step 17400, change in cost 0.000127044
step 17500, training accuracy 0.997901
step 17500, change in cost 0.000125458
step 17600, training accuracy 0.997901
step 17600, change in cost 0.000123989
step 17700, training accuracy 0.997901
step 17700, change in cost 0.000122531
step 17800, training accuracy 0.997901
step 17800, change in cost 0.000121033
step 17900, training accuracy 0.997901
step 17900, change in cost 0.000119603
step 18000, training accuracy 0.997901
step 18000, change in cost 0.000118217
step 18100, training accuracy 0.997901
step 18100, change in cost 0.000116779
step 18200, training accuracy 0.997901
step 18200, change in cost 0.000115403
step 18300, training accuracy 0.997901
step 18300, change in cost 0.000114088
step 18400, training accuracy 0.997901
step 18400, change in cost 0.000112717
step 18500, training accuracy 0.997901
step 18500, change in cost 0.000111395
step 18600, training accuracy 0.997901
step 18600, change in cost 0.000110127
step 18700, training accuracy 0.997901
step 18700, change in cost 0.000108781
step 18800, training accuracy 0.997901
step 18800, change in cost 0.000107568
step 18900, training accuracy 0.997901
step 18900, change in cost 0.000106321
step 19000, training accuracy 0.997901
step 19000, change in cost 0.000105033
step 19100, training accuracy 0.997901
step 19100, change in cost 0.000103892
step 19200, training accuracy 0.997901
step 19200, change in cost 0.000102638
step 19300, training accuracy 0.997901
step 19300, change in cost 0.000101475
step 19400, training accuracy 0.997901
step 19400, change in cost 0.000100307
step 19500, training accuracy 0.997901
step 19500, change in cost 9.91439e-05
change in cost 9.91439e-05; convergence.
final accuracy on test set: 0.990476
Error:  0.0020986199379






    




/usr/local/lib/python3.4/dist-packages/numpy/core/_methods.py:59: RuntimeWarning: Mean of empty slice.
  warnings.warn("Mean of empty slice.", RuntimeWarning)






    

In [6]:

    

errors[:]


    

    
Out[6]:





[0.53305351734161377,
 0.41448056697845459,
 0.33788037300109863,
 0.28541445732116699,
 0.25393491983413696,
 0.23294860124588013,
 0.21091288328170776,
 0.17943334579467773,
 0.16264426708221436,
 0.13011544942855835,
 0.11017841100692749,
 0.097586572170257568,
 0.29171037673950195,
 0.22980064153671265,
 0.21615952253341675,
 0.20251834392547607,
 0.17943334579467773,
 0.16474294662475586,
 0.14900314807891846,
 0.12696748971939087,
 0.11017841100692749,
 0.099685192108154297,
 0.08394545316696167,
 0.074501574039459229,
 0.054564535617828369,
 0.045120656490325928,
 0.040923416614532471,
 0.03462749719619751,
 0.031479537487030029,
 0.031479537487030029,
 0.031479537487030029,
 0.031479537487030029,
 0.031479537487030029,
 0.029380917549133301,
 0.028331577777862549,
 0.028331577777862549,
 0.027282238006591797,
 0.02623295783996582,
 0.02623295783996582,
 0.025183618068695068,
 0.025183618068695068,
 0.025183618068695068,
 0.024134337902069092,
 0.024134337902069092,
 0.02308499813079834,
 0.022035658359527588,
 0.022035658359527588,
 0.020986378192901611,
 0.019937038421630859,
 0.019937038421630859,
 0.018887698650360107,
 0.018887698650360107,
 0.015739798545837402,
 0.015739798545837402,
 0.015739798545837402,
 0.015739798545837402,
 0.013641119003295898,
 0.012591838836669922,
 0.012591838836669922,
 0.012591838836669922,
 0.012591838836669922,
 0.012591838836669922,
 0.01154249906539917,
 0.010493159294128418,
 0.010493159294128418,
 0.010493159294128418,
 0.010493159294128418,
 0.0094438791275024414,
 0.0094438791275024414,
 0.0094438791275024414,
 0.0094438791275024414,
 0.0094438791275024414,
 0.0083945393562316895,
 0.0083945393562316895,
 0.0083945393562316895,
 0.0083945393562316895,
 0.0083945393562316895,
 0.0073451995849609375,
 0.0073451995849609375,
 0.0073451995849609375,
 0.0073451995849609375,
 0.0073451995849609375,
 0.0073451995849609375,
 0.0062959194183349609,
 0.0062959194183349609,
 0.0062959194183349609,
 0.0062959194183349609,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.005246579647064209,
 0.0041972994804382324,
 0.0041972994804382324,
 0.0041972994804382324,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0031479597091674805,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285,
 0.0020986199378967285]

In [7]:

    

get_ipython().magic('matplotlib inline')
plt.title('Error Plot')
plt.plot([np.mean(errors[i-50:i]) for i in range(len(errors))])
plt.show()


    

    




/usr/local/lib/python3.4/dist-packages/numpy/core/_methods.py:59: RuntimeWarning: Mean of empty slice.
  warnings.warn("Mean of empty slice.", RuntimeWarning)






    

In [14]:

    




    

In [22]:

    




    

In [ ]: