notebook.community

Edit and run



In [1]:

    
import sys
sys.path.insert(0, '../models')



In [2]:

    
import pandas as pd
import numpy as np



In [3]:

    
from sklearn.decomposition import PCA



In [4]:

    
%matplotlib inline



In [5]:

    
import matplotlib.pyplot as plt



In [6]:

    
from UFCNN1 import draw_model, print_nodes_shapes, save_neuralnet, ufcnn_model_concat, ufcnn_model_sum, ufcnn_model, \
prepare_tradcom_classification, train_and_predict_classification, get_tradcom_normalization









    



Using TensorFlow backend.



In [7]:

    
from UFCNN1 import standardize_inputs, ufcnn_model_seq, print_layers_shapes



In [8]:

    
from keras.optimizers import SGD



In [9]:

    
sgd = SGD(lr=0.0005, decay=1e-6, momentum=0.9, nesterov=True)



In [10]:

    
model_concat = ufcnn_model_concat(regression = False, output_dim=3, features=32, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd )



In [11]:

    
print_nodes_shapes(model_concat)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
input_padding : <class 'keras.layers.convolutional.ZeroPadding1D'> : (None, 500, 32) : (None, 504, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 504, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 3)
activation : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)
output : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)



In [9]:

    
model_concat = ufcnn_model_concat(regression = False, output_dim=3, features=32, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd )



In [13]:

    
print_nodes_shapes(model_concat)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 3)
activation : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)
output : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)



In [11]:

    
model_concat_regression = ufcnn_model_concat(regression = True, output_dim=1, features=32, 
       loss="mse", sequence_length=500, optimizer=sgd )



In [12]:

    
print_nodes_shapes(model_concat_regression)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 1)
output : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 1)



In [9]:

    
model_concat = ufcnn_model(regression = False, output_dim=3, features=32, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd, mode='concat' )



In [10]:

    
print_nodes_shapes(model_concat)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 3)
activation : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)
output : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)



In [10]:

    
draw_model(model_concat)



In [ ]:

    
model_sum = ufcnn_model_sum(regression = False, output_dim=3, features=32, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd, mode='sum' )



In [14]:

    
model_sum = ufcnn_model(regression = False, output_dim=3, features=32, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd, mode='sum' )



In [15]:

    
print_nodes_shapes(model_sum)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu7 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
flatten : <class 'keras.layers.core.Flatten'> : (None, 500, 150) : (None, 75000)
dense : <class 'keras.layers.core.Dense'> : (None, 75000) : (None, 3)
output : <class 'keras.layers.core.Dense'> : (None, 75000) : (None, 3)



In [18]:

    
print_nodes_shapes(model_sum)









    



input : <class 'keras.layers.core.Layer'> : (None, 500, 32) : (None, 500, 32)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 32) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu7 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
flatten : <class 'keras.layers.core.Flatten'> : (None, 500, 150) : (None, 75000)
dense : <class 'keras.layers.core.Dense'> : (None, 75000) : (None, 3)
output : <class 'keras.layers.core.Dense'> : (None, 75000) : (None, 3)



In [30]:

    
Xdf = pd.read_csv('../models/training_data_large/prod_data_20130729v.txt', sep=" ", index_col = 0, header = None,)



In [12]:

    
ydf = pd.read_csv('../models/training_data_large/signal_20130729v.csv', index_col = 0, names = ['signal',], )



In [13]:

    
ydf









    Out[13]:






  
    
      
      signal
    
  
  
    
      43264234
      -1
    
    
      43264739
      -1
    
    
      43265282
      -1
    
    
      43265789
      -1
    
    
      43266318
      -1
    
    
      43266846
      -1
    
    
      43267353
      0
    
    
      43267872
      0
    
    
      43268387
      0
    
    
      43268898
      0
    
    
      43269621
      0
    
    
      43270209
      0
    
    
      43270836
      0
    
    
      43271382
      0
    
    
      43271922
      0
    
    
      43272451
      0
    
    
      43273085
      -1
    
    
      43273598
      -1
    
    
      43274160
      -1
    
    
      43274691
      -1
    
    
      43275289
      -1
    
    
      43275808
      -1
    
    
      43276380
      -1
    
    
      43276926
      -1
    
    
      43277505
      -1
    
    
      43278047
      -1
    
    
      43278688
      -1
    
    
      43279255
      -1
    
    
      43279761
      -1
    
    
      43280684
      -1
    
    
      ...
      ...
    
    
      71985083
      1
    
    
      71985584
      1
    
    
      71986088
      1
    
    
      71986619
      1
    
    
      71987123
      1
    
    
      71987635
      1
    
    
      71988149
      1
    
    
      71988656
      1
    
    
      71989176
      1
    
    
      71989685
      1
    
    
      71990186
      1
    
    
      71990689
      1
    
    
      71991192
      1
    
    
      71991702
      1
    
    
      71992205
      1
    
    
      71992706
      1
    
    
      71993207
      1
    
    
      71993715
      1
    
    
      71994219
      1
    
    
      71994722
      1
    
    
      71995233
      1
    
    
      71995735
      1
    
    
      71996237
      1
    
    
      71996747
      1
    
    
      71997248
      1
    
    
      71997749
      1
    
    
      71998250
      1
    
    
      71998751
      1
    
    
      71999257
      1
    
    
      71999758
      1
    
  

53353 rows × 1 columns



In [14]:

    
Xdf.describe()









    Out[14]:






  
    
      
      1
      2
      3
      4
      5
      6
      7
      8
      9
      10
      ...
      23
      24
      25
      26
      27
      28
      29
      30
      31
      32
    
  
  
    
      count
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      ...
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      53353.000000
    
    
      mean
      4528.503695
      4527.986467
      292.032126
      4529.025566
      314.730943
      -0.006248
      0.061189
      0.063878
      -0.056281
      -0.056014
      ...
      -68.085227
      -0.048457
      0.040879
      -0.005342
      -0.175384
      -0.074360
      -0.004406
      0.205098
      -0.074945
      0.002297
    
    
      std
      7.091962
      7.096300
      402.466606
      7.092489
      268.652928
      0.969723
      0.501643
      0.979621
      0.612950
      0.994605
      ...
      346.127357
      16.470564
      0.709954
      0.327163
      1.599705
      0.747817
      0.231498
      1.720832
      2.004901
      0.189858
    
    
      min
      4513.442966
      4513.000000
      10.000000
      4514.000000
      10.000000
      -5.300362
      -3.142692
      -5.165990
      -3.458124
      -7.612156
      ...
      -2170.000000
      -68.912082
      -2.860846
      -3.105852
      -6.433072
      -5.428968
      -0.827306
      -4.982010
      -15.306670
      -1.272888
    
    
      25%
      4523.483634
      4523.000000
      130.000000
      4524.000000
      120.000000
      -0.503704
      -0.134776
      -0.398840
      -0.296592
      -0.583548
      ...
      -228.888888
      -10.634058
      -0.433480
      -0.113374
      -1.130916
      -0.356486
      -0.167236
      -1.100764
      -0.944650
      0.000000
    
    
      50%
      4528.321634
      4528.000000
      220.000000
      4529.000000
      240.000000
      0.022336
      0.000000
      0.030460
      -0.027716
      -0.048040
      ...
      -76.666666
      0.521168
      0.048912
      0.000000
      -0.217928
      -0.014408
      0.000384
      0.163566
      -0.000000
      0.000000
    
    
      75%
      4533.542640
      4533.000000
      360.000000
      4534.000000
      450.000000
      0.520892
      0.233978
      0.458288
      0.194748
      0.461766
      ...
      97.777778
      11.078318
      0.507192
      0.109450
      0.750900
      0.278332
      0.166794
      1.533210
      0.832284
      0.000000
    
    
      max
      4544.511066
      4544.000000
      7430.000000
      4545.000000
      2570.000000
      5.462836
      3.951678
      6.683222
      3.634796
      5.598256
      ...
      4170.000000
      74.140150
      4.666812
      3.842634
      6.560304
      3.466916
      0.787244
      7.748302
      10.165278
      1.104922
    
  

8 rows × 32 columns



In [11]:

    
indicators = Xdf[range(6,33)]



In [15]:

    
indicators = indicators.sub(indicators.mean())



In [16]:

    
indicators = indicators.div(indicators.std())



In [17]:

    
indicators.describe()









    Out[17]:






  
    
      
      6
      7
      8
      9
      10
      11
      12
      13
      14
      15
      ...
      23
      24
      25
      26
      27
      28
      29
      30
      31
      32
    
  
  
    
      count
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      ...
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
    
    
      mean
      1.779253e-16
      -5.794826e-15
      1.576823e-16
      -1.959310e-15
      -2.929908e-16
      -6.605612e-17
      -6.967003e-14
      6.179443e-16
      1.230348e-14
      -2.336469e-15
      ...
      4.219068e-15
      3.036451e-17
      1.497982e-15
      -7.750939e-17
      8.523370e-17
      -6.285985e-17
      2.706170e-16
      4.421498e-15
      1.688693e-16
      1.054767e-16
    
    
      std
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      ...
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
    
    
      min
      -5.459411e+00
      -6.386769e+00
      -5.338667e+00
      -5.549951e+00
      -7.597130e+00
      -3.461954e+00
      -3.017210e+00
      -3.889494e+00
      -3.361662e+00
      -4.148984e+00
      ...
      -6.072663e+00
      -4.181012e+00
      -4.087199e+00
      -9.476943e+00
      -3.911775e+00
      -7.160319e+00
      -3.554669e+00
      -3.014303e+00
      -7.597245e+00
      -6.716526e+00
    
    
      25%
      -5.129881e-01
      -3.906464e-01
      -4.723444e-01
      -3.920565e-01
      -5.303954e-01
      -7.493184e-01
      -6.829345e-01
      -7.005503e-01
      -7.471192e-01
      -7.821417e-01
      ...
      -4.645795e-01
      -6.426981e-01
      -6.681542e-01
      -3.302073e-01
      -5.973172e-01
      -3.772660e-01
      -7.033728e-01
      -7.588551e-01
      -4.337893e-01
      -1.210090e-02
    
    
      50%
      2.947630e-02
      -1.219775e-01
      -3.411349e-02
      4.660233e-02
      8.017472e-03
      -4.055363e-02
      2.152361e-02
      7.690110e-02
      1.424535e-02
      -1.083107e-01
      ...
      -2.479272e-02
      3.458445e-02
      1.131478e-02
      1.632894e-02
      -2.659460e-02
      8.016953e-02
      2.069328e-02
      -2.413491e-02
      3.738106e-02
      -1.210090e-02
    
    
      75%
      5.435986e-01
      3.444454e-01
      4.026148e-01
      4.095421e-01
      5.205889e-01
      7.709655e-01
      4.731585e-01
      7.280373e-01
      6.986443e-01
      7.757446e-01
      ...
      4.791965e-01
      6.755552e-01
      6.568210e-01
      3.508712e-01
      5.790344e-01
      4.716290e-01
      7.395326e-01
      7.717848e-01
      4.525058e-01
      -1.210090e-02
    
    
      max
      5.639844e+00
      7.755486e+00
      6.757049e+00
      6.021822e+00
      5.684942e+00
      3.056568e+00
      3.994932e+00
      3.284439e+00
      3.866930e+00
      3.173313e+00
      ...
      1.224429e+01
      4.504315e+00
      6.515817e+00
      1.176163e+01
      4.210581e+00
      4.735485e+00
      3.419683e+00
      4.383464e+00
      5.107595e+00
      5.807631e+00
    
  

8 rows × 27 columns



In [53]:

    
plt.figure(figsize=(18,12))
plt.plot(indicators[-5000:][[6, 7, 8, 9, 10]])









    Out[53]:





[<matplotlib.lines.Line2D at 0x7f0324658490>,
 <matplotlib.lines.Line2D at 0x7f0324658610>,
 <matplotlib.lines.Line2D at 0x7f0324658710>,
 <matplotlib.lines.Line2D at 0x7f0324658850>,
 <matplotlib.lines.Line2D at 0x7f0324658990>]



In [40]:

    
pca = PCA(n_components=15)
pca.fit(indicators.values)









    Out[40]:





PCA(copy=True, n_components=15, whiten=False)



In [41]:

    
pca.explained_variance_ratio_









    Out[41]:





array([ 0.20578496,  0.13508374,  0.09762137,  0.07507347,  0.06276689,
        0.05787578,  0.04907913,  0.04143535,  0.03605835,  0.0351595 ,
        0.03143687,  0.02887315,  0.02437674,  0.02275628,  0.01857632])



In [33]:

    
pca.explained_variance_









    Out[33]:





array([ 5.55608985,  3.64719274,  2.63572761,  2.02694562,  1.69467438])



In [42]:

    
np.sum(pca.explained_variance_ratio_)









    Out[42]:





0.9219579116377834



In [52]:

    
plt.figure(figsize=(18,12))
plt.plot(pca_ind[-5000:, 0:5])









    Out[52]:





[<matplotlib.lines.Line2D at 0x7f032473fb50>,
 <matplotlib.lines.Line2D at 0x7f032473fcd0>,
 <matplotlib.lines.Line2D at 0x7f032473fdd0>,
 <matplotlib.lines.Line2D at 0x7f032473ff10>,
 <matplotlib.lines.Line2D at 0x7f032474f090>]



In [44]:

    
pca.components_.shape









    Out[44]:





(15, 27)



In [45]:

    
pca_ind = pca.transform(indicators.values)



In [79]:

    
pca_ind_df = pd.DataFrame(data=pca_ind, index=indicators.index)



In [78]:

    
Xdf.shape









    Out[78]:





(53353, 32)



In [121]:

    
plt.figure(figsize=(18,12))
#plt.plot(Xdf[0:10000][[2, 4]])
#plt.plot(ydf.add(4516)[0:10000])
plt.plot(ba_df_std.add(4520)[0:10000][[2, 4]])
#plt.plot(Xdf_std.add(4520)[0:10000][[2, 4]])
#plt.plot(Xdf.div(Xdf.std()).add(4510)[0:10000][[3, 5]])
#plt.plot(pca_ind_df.add(4505)[0:10000][[0,]])
#plt.plot(indicators.add(4505)[0:10000][[6,]])









    Out[121]:





[<matplotlib.lines.Line2D at 0x7f0306481110>,
 <matplotlib.lines.Line2D at 0x7f0306481250>]



In [110]:

    
plt.figure(figsize=(18,12))
plt.plot(Xdf[0:1000][[2, 4]])
plt.plot(ydf.add(4512)[0:1000])









    Out[110]:





[<matplotlib.lines.Line2D at 0x7f0306d55210>]



In [97]:

    
Xdf_std = Xdf.sub(Xdf.mean())

    Xdf_std = Xdf_std.div(Xdf_std.std())



In [107]:

    
ba_df_std = Xdf[[2, 4]].sub(Xdf[[2, 4]].mean(axis=1).mean())

    ba_df_std = ba_df_std.div(ba_df_std.std())



In [108]:

    
ba_df_std.describe()









    Out[108]:






  
    
      
      2
      4
    
  
  
    
      count
      53353.000000
      53353.000000
    
    
      mean
      -0.073214
      0.073253
    
    
      std
      1.000000
      1.000000
    
    
      min
      -2.185085
      -2.045265
    
    
      25%
      -0.775900
      -0.635322
    
    
      50%
      -0.071307
      0.069649
    
    
      75%
      0.633285
      0.774620
    
    
      max
      2.183389
      2.325556



In [99]:

    
Xdf_std.describe()









    Out[99]:






  
    
      
      1
      2
      3
      4
      5
      6
      7
      8
      9
      10
      ...
      23
      24
      25
      26
      27
      28
      29
      30
      31
      32
    
  
  
    
      count
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      ...
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
    
    
      mean
      -1.859223e-11
      3.492877e-14
      -2.397198e-17
      -2.471777e-15
      8.629912e-17
      1.779253e-16
      -5.794826e-15
      1.576823e-16
      -1.959310e-15
      -2.929908e-16
      ...
      4.219068e-15
      3.036451e-17
      1.497982e-15
      -7.750939e-17
      8.523370e-17
      -6.285985e-17
      2.706170e-16
      4.421498e-15
      1.688693e-16
      1.054767e-16
    
    
      std
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      ...
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
    
    
      min
      -2.123634e+00
      -2.111871e+00
      -7.007591e-01
      -2.118518e+00
      -1.134292e+00
      -5.459411e+00
      -6.386769e+00
      -5.338667e+00
      -5.549951e+00
      -7.597130e+00
      ...
      -6.072663e+00
      -4.181012e+00
      -4.087199e+00
      -9.476943e+00
      -3.911775e+00
      -7.160319e+00
      -3.554669e+00
      -3.014303e+00
      -7.597245e+00
      -6.716526e+00
    
    
      25%
      -7.078522e-01
      -7.026855e-01
      -4.025977e-01
      -7.085757e-01
      -7.248421e-01
      -5.129881e-01
      -3.906464e-01
      -4.723444e-01
      -3.920565e-01
      -5.303954e-01
      ...
      -4.645795e-01
      -6.426981e-01
      -6.681542e-01
      -3.302073e-01
      -5.973172e-01
      -3.772660e-01
      -7.033728e-01
      -7.588551e-01
      -4.337893e-01
      -1.210090e-02
    
    
      50%
      -2.567144e-02
      1.906981e-03
      -1.789767e-01
      -3.604598e-03
      -2.781691e-01
      2.947630e-02
      -1.219775e-01
      -3.411349e-02
      4.660233e-02
      8.017472e-03
      ...
      -2.479272e-02
      3.458445e-02
      1.131478e-02
      1.632894e-02
      -2.659460e-02
      8.016953e-02
      2.069328e-02
      -2.413491e-02
      3.738106e-02
      -1.210090e-02
    
    
      75%
      7.105150e-01
      7.064995e-01
      1.688783e-01
      7.013665e-01
      5.035086e-01
      5.435986e-01
      3.444454e-01
      4.026148e-01
      4.095421e-01
      5.205889e-01
      ...
      4.791965e-01
      6.755552e-01
      6.568210e-01
      3.508712e-01
      5.790344e-01
      4.716290e-01
      7.395326e-01
      7.717848e-01
      4.525058e-01
      -1.210090e-02
    
    
      max
      2.257115e+00
      2.256603e+00
      1.773555e+01
      2.252303e+00
      8.394731e+00
      5.639844e+00
      7.755486e+00
      6.757049e+00
      6.021822e+00
      5.684942e+00
      ...
      1.224429e+01
      4.504315e+00
      6.515817e+00
      1.176163e+01
      4.210581e+00
      4.735485e+00
      3.419683e+00
      4.383464e+00
      5.107595e+00
      5.807631e+00
    
  

8 rows × 32 columns



In [15]:

    
(mean, std) = get_tradcom_normalization(filename = '../models/training_data_large/prod_data_20130729v.txt', mean = None, std = None)









    



Result after max 1     4529.025566
2     4529.025566
3      314.730943
4     4529.025566
5      314.730943
6       -0.006248
7        0.061189
8        0.063878
9       -0.056281
10      -0.056014
11      -0.052083
12       0.461357
13       0.436503
14      -0.433617
15      -0.492630
16      -0.030645
17       1.211221
18       1.358510
19      -1.141625
20      -1.358383
21      -0.013793
22      -1.218955
23     -68.085227
24      -0.048457
25       0.040879
26      -0.005342
27      -0.175384
28      -0.074360
29      -0.004406
30       0.205098
31      -0.074945
32       0.002297
dtype: float64
Mean Loc
1     4529.025566
2     4529.025566
3      314.730943
4     4529.025566
5      314.730943
6       -0.006248
7        0.061189
8        0.063878
9       -0.056281
10      -0.056014
11      -0.052083
12       0.461357
13       0.436503
14      -0.433617
15      -0.492630
16      -0.030645
17       1.211221
18       1.358510
19      -1.141625
20      -1.358383
21      -0.013793
22      -1.218955
23     -68.085227
24      -0.048457
25       0.040879
26      -0.005342
27      -0.175384
28      -0.074360
29      -0.004406
30       0.205098
31      -0.074945
32       0.002297
dtype: float64
Mean
1     4529.025566
2     4529.025566
3      314.730943
4     4529.025566
5      314.730943
6       -0.006248
7        0.061189
8        0.063878
9       -0.056281
10      -0.056014
11      -0.052083
12       0.461357
13       0.436503
14      -0.433617
15      -0.492630
16      -0.030645
17       1.211221
18       1.358510
19      -1.141625
20      -1.358383
21      -0.013793
22      -1.218955
23     -68.085227
24      -0.048457
25       0.040879
26      -0.005342
27      -0.175384
28      -0.074360
29      -0.004406
30       0.205098
31      -0.074945
32       0.002297
dtype: float64
Result after max 1       7.096300
2       7.096300
3     402.466606
4       7.096300
5     402.466606
6       0.969723
7       0.501643
8       0.979621
9       0.612950
10      0.994605
11      2.246582
12      1.355967
13      2.150650
14      1.693381
15      2.241740
16      3.657930
17      2.168240
18      3.282834
19      2.608395
20      3.504068
21      1.526378
22      0.685365
23    346.127357
24     16.470564
25      0.709954
26      0.327163
27      1.599705
28      0.747817
29      0.231498
30      1.720832
31      2.004901
32      0.189858
dtype: float64
Std Loc
1       7.096300
2       7.096300
3     402.466606
4       7.096300
5     402.466606
6       0.969723
7       0.501643
8       0.979621
9       0.612950
10      0.994605
11      2.246582
12      1.355967
13      2.150650
14      1.693381
15      2.241740
16      3.657930
17      2.168240
18      3.282834
19      2.608395
20      3.504068
21      1.526378
22      0.685365
23    346.127357
24     16.470564
25      0.709954
26      0.327163
27      1.599705
28      0.747817
29      0.231498
30      1.720832
31      2.004901
32      0.189858
dtype: float64
Std
1       7.096300
2       7.096300
3     402.466606
4       7.096300
5     402.466606
6       0.969723
7       0.501643
8       0.979621
9       0.612950
10      0.994605
11      2.246582
12      1.355967
13      2.150650
14      1.693381
15      2.241740
16      3.657930
17      2.168240
18      3.282834
19      2.608395
20      3.504068
21      1.526378
22      0.685365
23    346.127357
24     16.470564
25      0.709954
26      0.327163
27      1.599705
28      0.747817
29      0.231498
30      1.720832
31      2.004901
32      0.189858
dtype: float64



In [34]:

    
ba_std = standardize_columns(Xdf[[2, 4]])



In [35]:

    
ba_std.describe()









    Out[35]:






  
    
      
      2
      4
    
  
  
    
      count
      53353.000000
      53353.000000
    
    
      mean
      -0.073039
      0.073039
    
    
      std
      0.997606
      0.997071
    
    
      min
      -2.179854
      -2.039273
    
    
      25%
      -0.774042
      -0.633461
    
    
      50%
      -0.071136
      0.069445
    
    
      75%
      0.631770
      0.772351
    
    
      max
      2.178163
      2.318744



In [35]:

    
Xdf_std = standardize_inputs(source='../models/training_data_large/prod_data_20130729v.txt', colgroups = [[2, 4], [3, 5]])









    



(53353, 2)
(53353, 4)
[1, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32]
(53353, 32)



In [36]:

    
Xdf_std = standardize_inputs(source=Xdf, colgroups = [[2, 4], [3, 5]])









    



(53353, 2)
(53353, 4)
[1, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32]
(53353, 32)



In [15]:

    
Xdf_std.describe()









    Out[15]:






  
    
      
      2
      4
      3
      5
      1
      6
      7
      8
      9
      10
      ...
      23
      24
      25
      26
      27
      28
      29
      30
      31
      32
    
  
  
    
      count
      53353.000000
      53353.000000
      53353.000000
      53353.000000
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      ...
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
      5.335300e+04
    
    
      mean
      -0.073039
      0.073039
      -0.033151
      0.033151
      -1.859220e-11
      1.843179e-16
      -5.769256e-15
      1.619440e-16
      -1.958244e-15
      -2.929908e-16
      ...
      4.236115e-15
      3.036451e-17
      1.497982e-15
      -7.750939e-17
      1.033459e-16
      -6.285985e-17
      2.706170e-16
      4.430021e-15
      1.725982e-16
      1.044113e-16
    
    
      std
      0.997606
      0.997071
      1.175600
      0.784732
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      ...
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
      1.000000e+00
    
    
      min
      -2.179854
      -2.039273
      -0.856964
      -0.856964
      -2.123634e+00
      -5.459411e+00
      -6.386769e+00
      -5.338667e+00
      -5.549951e+00
      -7.597130e+00
      ...
      -6.072663e+00
      -4.181012e+00
      -4.087199e+00
      -9.476943e+00
      -3.911775e+00
      -7.160319e+00
      -3.554669e+00
      -3.014303e+00
      -7.597245e+00
      -6.716526e+00
    
    
      25%
      -0.774042
      -0.633461
      -0.506445
      -0.535655
      -7.078522e-01
      -5.129881e-01
      -3.906464e-01
      -4.723444e-01
      -3.920565e-01
      -5.303954e-01
      ...
      -4.645795e-01
      -6.426981e-01
      -6.681542e-01
      -3.302073e-01
      -5.973172e-01
      -3.772660e-01
      -7.033728e-01
      -7.588551e-01
      -4.337893e-01
      -1.210090e-02
    
    
      50%
      -0.071136
      0.069445
      -0.243556
      -0.185137
      -2.567144e-02
      2.947630e-02
      -1.219775e-01
      -3.411349e-02
      4.660233e-02
      8.017472e-03
      ...
      -2.479272e-02
      3.458445e-02
      1.131478e-02
      1.632894e-02
      -2.659460e-02
      8.016953e-02
      2.069328e-02
      -2.413491e-02
      3.738106e-02
      -1.210090e-02
    
    
      75%
      0.631770
      0.772351
      0.165382
      0.428271
      7.105150e-01
      5.435986e-01
      3.444454e-01
      4.026148e-01
      4.095421e-01
      5.205889e-01
      ...
      4.791965e-01
      6.755552e-01
      6.568210e-01
      3.508712e-01
      5.790344e-01
      4.716290e-01
      7.395326e-01
      7.717848e-01
      4.525058e-01
      -1.210090e-02
    
    
      max
      2.178163
      2.318744
      20.816764
      6.620764
      2.257115e+00
      5.639844e+00
      7.755486e+00
      6.757049e+00
      6.021822e+00
      5.684942e+00
      ...
      1.224429e+01
      4.504315e+00
      6.515817e+00
      1.176163e+01
      4.210581e+00
      4.735485e+00
      3.419683e+00
      4.383464e+00
      5.107595e+00
      5.807631e+00
    
  

8 rows × 32 columns



In [18]:

    
type(Xdf_std)









    Out[18]:





pandas.core.frame.DataFrame



In [15]:

    
(X, y) = prepare_tradcom_classification(training = True, sequence_length = 500, features = 4, output_dim = 3, filename='../models/training_data_large/prod_data_20130729v.txt')









    



Creating files  ../models/training_data_large/prod_data_20130729v_X.npy  and  ../models/training_data_large/prod_data_20130729v_y.npy
Working on Input files:  ../models/training_data_large/prod_data_20130729v.txt ,  ../models/training_data_large/signal_20130729v.csv
(53353, 2)
(53353, 4)
[]
(53353, 4)
                  2             4             3             5
count  53353.000000  53353.000000  53353.000000  53353.000000
mean      -0.073039      0.073039     -0.033151      0.033151
std        0.997606      0.997071      1.175600      0.784732
min       -2.179854     -2.039273     -0.856964     -0.856964
25%       -0.774042     -0.633461     -0.506445     -0.535655
50%       -0.071136      0.069445     -0.243556     -0.185137
75%        0.631770      0.772351      0.165382      0.428271
max        2.178163      2.318744     20.816764      6.620764
(52854, 500, 4)
              2         4         3         5
0                                            
71997749  1.897  2.037581 -0.564865  0.428271
71998250  1.897  2.037581 -0.564865  0.428271
71998751  1.897  2.037581 -0.564865  0.399061
71999257  1.897  2.037581 -0.068297  0.369851
71999758  1.897  2.037581 -0.068297  0.369851
(52854, 500, 3)



In [17]:

    
plt.figure(figsize=(18,6))
    plt.plot(X[-24000][:, 0:2]+3)
    plt.plot(y[-24000][:, 0:2])









    Out[17]:





[<matplotlib.lines.Line2D at 0x7fecabbe1150>,
 <matplotlib.lines.Line2D at 0x7febfb440590>]



In [11]:

    
(X, y) = prepare_tradcom_classification(training = True, sequence_length = 5000, features = 4, output_dim = 3, filename='../models/training_data_large/prod_data_20130729v.txt')









    



Creating files  ../models/training_data_large/prod_data_20130729v_X.npy  and  ../models/training_data_large/prod_data_20130729v_y.npy
Working on Input files:  ../models/training_data_large/prod_data_20130729v.txt ,  ../models/training_data_large/signal_20130729v.csv
(53353, 2)
(53353, 4)
[]
(53353, 4)
                  2             4             3             5
count  53353.000000  53353.000000  53353.000000  53353.000000
mean      -0.073039      0.073039     -0.033151      0.033151
std        0.997606      0.997071      1.175600      0.784732
min       -2.179854     -2.039273     -0.856964     -0.856964
25%       -0.774042     -0.633461     -0.506445     -0.535655
50%       -0.071136      0.069445     -0.243556     -0.185137
75%        0.631770      0.772351      0.165382      0.428271
max        2.178163      2.318744     20.816764      6.620764
(48354, 5000, 4)
              2         4         3         5
0                                            
71997749  1.897  2.037581 -0.564865  0.428271
71998250  1.897  2.037581 -0.564865  0.428271
71998751  1.897  2.037581 -0.564865  0.399061
71999257  1.897  2.037581 -0.068297  0.369851
71999758  1.897  2.037581 -0.068297  0.369851
(48354, 5000, 3)



In [13]:

    
plt.figure(figsize=(18,6))
    plt.plot(X[-20000][:, 0:2]+3)
    plt.plot(y[-20000][:, 0:2])









    Out[13]:





[<matplotlib.lines.Line2D at 0x7fe9828e8350>,
 <matplotlib.lines.Line2D at 0x7fecabc67d10>]



In [12]:

    
(X, y) = prepare_tradcom_classification(training = True, stack=False, sequence_length = 500, features = 4, output_dim = 3, filename='../models/training_data_large/prod_data_20130729v.txt')









    



Found files  ../models/training_data_large/prod_data_20130729v_X.npy  and  ../models/training_data_large/prod_data_20130729v_y.npy



In [29]:

    
X.shape









    Out[29]:





(53353, 4)



In [13]:

    
X = X.reshape((1, X.shape[0], X.shape[1]))



In [14]:

    
X.shape









    Out[14]:





(1, 53353, 4)



In [30]:

    
y.shape









    Out[30]:





(53353, 3)



In [15]:

    
y = y.reshape((1, y.shape[0], y.shape[1]))



In [16]:

    
y.shape









    Out[16]:





(1, 53353, 3)



In [10]:

    
model = ufcnn_model_concat(day_data_length=50000, regression = False, output_dim=3, features=4, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd )



In [11]:

    
print_nodes_shapes(model)









    



input : <class 'keras.layers.core.Layer'> : (None, None, 4) : (None, None, 4)
input_padding : <class 'keras.layers.convolutional.ZeroPadding1D'> : (None, None, 4) : (None, None, 4)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 4) : (None, None, 150)
relu1 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu2 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu3 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu4 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 300) : (None, None, 150)
relu5 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 300) : (None, None, 150)
relu6 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 3)
activation : <class 'keras.layers.core.Activation'> : (None, None, 3) : (None, None, 3)
output : <class 'keras.layers.core.Activation'> : (None, None, 3) : (None, None, 3)



In [17]:

    
history = model.fit({'input': X, 'output': y},
                      verbose=2,
                      nb_epoch=1,
                      shuffle=False,
                      batch_size=1)









    



Epoch 1/1
18s - loss: 1.0195



In [19]:

    
history.params









    Out[19]:





{'batch_size': 1,
 'do_validation': False,
 'metrics': ['loss', 'val_loss'],
 'nb_epoch': 1,
 'nb_sample': 1,
 'verbose': 2}



In [20]:

    
print_nodes_shapes(model)









    



input : <class 'keras.layers.core.Layer'> : (None, None, 4) : (None, None, 4)
input_padding : <class 'keras.layers.convolutional.ZeroPadding1D'> : (None, None, 4) : (None, None, 4)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 4) : (None, None, 150)
relu1 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu2 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu3 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 150)
relu4 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 300) : (None, None, 150)
relu5 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 300) : (None, None, 150)
relu6 : <class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 3)
activation : <class 'keras.layers.core.Activation'> : (None, None, 3) : (None, None, 3)
output : <class 'keras.layers.core.Activation'> : (None, None, 3) : (None, None, 3)



In [16]:

    
model = ufcnn_model_seq(regression = False, output_dim=3, features=4, 
       loss="categorical_crossentropy", sequence_length=500, optimizer=sgd )



In [17]:

    
print_layers_shapes(model)









    



<class 'keras.layers.convolutional.ZeroPadding1D'> : (None, None, 4) : (None, None, 4)
<class 'keras.layers.convolutional.Convolution1D'> : (None, None, 4) : (None, None, 150)
<class 'keras.layers.core.Activation'> : (None, None, 150) : (None, None, 150)
<class 'keras.layers.convolutional.Convolution1D'> : (None, None, 150) : (None, None, 3)
<class 'keras.layers.core.Activation'> : (None, None, 3) : (None, None, 3)



In [23]:

    
history = model.fit(X, y,
                      verbose=2,
                      nb_epoch=1,
                      batch_size=1
                    )









    



Epoch 1/1
14s - loss: 1.1141



In [ ]:

	signal
43264234	-1
43264739	-1
43265282	-1
43265789	-1
43266318	-1
43266846	-1
43267353	0
43267872	0
43268387	0
43268898	0
43269621	0
43270209	0
43270836	0
43271382	0
43271922	0
43272451	0
43273085	-1
43273598	-1
43274160	-1
43274691	-1
43275289	-1
43275808	-1
43276380	-1
43276926	-1
43277505	-1
43278047	-1
43278688	-1
43279255	-1
43279761	-1
43280684	-1
...	...
71985083	1
71985584	1
71986088	1
71986619	1
71987123	1
71987635	1
71988149	1
71988656	1
71989176	1
71989685	1
71990186	1
71990689	1
71991192	1
71991702	1
71992205	1
71992706	1
71993207	1
71993715	1
71994219	1
71994722	1
71995233	1
71995735	1
71996237	1
71996747	1
71997248	1
71997749	1
71998250	1
71998751	1
71999257	1
71999758	1

	1	2	3	4	5	6	7	8	9	10	...	23	24	25	26	27	28	29	30	31	32
count	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	...	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000	53353.000000
mean	4528.503695	4527.986467	292.032126	4529.025566	314.730943	-0.006248	0.061189	0.063878	-0.056281	-0.056014	...	-68.085227	-0.048457	0.040879	-0.005342	-0.175384	-0.074360	-0.004406	0.205098	-0.074945	0.002297
std	7.091962	7.096300	402.466606	7.092489	268.652928	0.969723	0.501643	0.979621	0.612950	0.994605	...	346.127357	16.470564	0.709954	0.327163	1.599705	0.747817	0.231498	1.720832	2.004901	0.189858
min	4513.442966	4513.000000	10.000000	4514.000000	10.000000	-5.300362	-3.142692	-5.165990	-3.458124	-7.612156	...	-2170.000000	-68.912082	-2.860846	-3.105852	-6.433072	-5.428968	-0.827306	-4.982010	-15.306670	-1.272888
25%	4523.483634	4523.000000	130.000000	4524.000000	120.000000	-0.503704	-0.134776	-0.398840	-0.296592	-0.583548	...	-228.888888	-10.634058	-0.433480	-0.113374	-1.130916	-0.356486	-0.167236	-1.100764	-0.944650	0.000000
50%	4528.321634	4528.000000	220.000000	4529.000000	240.000000	0.022336	0.000000	0.030460	-0.027716	-0.048040	...	-76.666666	0.521168	0.048912	0.000000	-0.217928	-0.014408	0.000384	0.163566	-0.000000	0.000000
75%	4533.542640	4533.000000	360.000000	4534.000000	450.000000	0.520892	0.233978	0.458288	0.194748	0.461766	...	97.777778	11.078318	0.507192	0.109450	0.750900	0.278332	0.166794	1.533210	0.832284	0.000000
max	4544.511066	4544.000000	7430.000000	4545.000000	2570.000000	5.462836	3.951678	6.683222	3.634796	5.598256	...	4170.000000	74.140150	4.666812	3.842634	6.560304	3.466916	0.787244	7.748302	10.165278	1.104922

	6	7	8	9	10	11	12	13	14	15	...	23	24	25	26	27	28	29	30	31	32
count	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	...	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04	5.335300e+04
mean	1.779253e-16	-5.794826e-15	1.576823e-16	-1.959310e-15	-2.929908e-16	-6.605612e-17	-6.967003e-14	6.179443e-16	1.230348e-14	-2.336469e-15	...	4.219068e-15	3.036451e-17	1.497982e-15	-7.750939e-17	8.523370e-17	-6.285985e-17	2.706170e-16	4.421498e-15	1.688693e-16	1.054767e-16
std	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	...	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00	1.000000e+00
min	-5.459411e+00	-6.386769e+00	-5.338667e+00	-5.549951e+00	-7.597130e+00	-3.461954e+00	-3.017210e+00	-3.889494e+00	-3.361662e+00	-4.148984e+00	...	-6.072663e+00	-4.181012e+00	-4.087199e+00	-9.476943e+00	-3.911775e+00	-7.160319e+00	-3.554669e+00	-3.014303e+00	-7.597245e+00	-6.716526e+00
25%	-5.129881e-01	-3.906464e-01	-4.723444e-01	-3.920565e-01	-5.303954e-01	-7.493184e-01	-6.829345e-01	-7.005503e-01	-7.471192e-01	-7.821417e-01	...	-4.645795e-01	-6.426981e-01	-6.681542e-01	-3.302073e-01	-5.973172e-01	-3.772660e-01	-7.033728e-01	-7.588551e-01	-4.337893e-01	-1.210090e-02
50%	2.947630e-02	-1.219775e-01	-3.411349e-02	4.660233e-02	8.017472e-03	-4.055363e-02	2.152361e-02	7.690110e-02	1.424535e-02	-1.083107e-01	...	-2.479272e-02	3.458445e-02	1.131478e-02	1.632894e-02	-2.659460e-02	8.016953e-02	2.069328e-02	-2.413491e-02	3.738106e-02	-1.210090e-02
75%	5.435986e-01	3.444454e-01	4.026148e-01	4.095421e-01	5.205889e-01	7.709655e-01	4.731585e-01	7.280373e-01	6.986443e-01	7.757446e-01	...	4.791965e-01	6.755552e-01	6.568210e-01	3.508712e-01	5.790344e-01	4.716290e-01	7.395326e-01	7.717848e-01	4.525058e-01	-1.210090e-02
max	5.639844e+00	7.755486e+00	6.757049e+00	6.021822e+00	5.684942e+00	3.056568e+00	3.994932e+00	3.284439e+00	3.866930e+00	3.173313e+00	...	1.224429e+01	4.504315e+00	6.515817e+00	1.176163e+01	4.210581e+00	4.735485e+00	3.419683e+00	4.383464e+00	5.107595e+00	5.807631e+00