In [1]:

    

import pandas as pd
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
# from common import *


    

In [2]:

    

pred = np.load('pred_X_1_model_0_2017.npy')


    

In [5]:

    

# s = np.sum(pred, axis=1, keepdims=True )
# pred = pred * 2 / s


    

In [3]:

    

pred_original = np.load('pred_par0_2016.npy')


    

In [4]:

    

print(pred.shape)
print(pred_original.shape)


    

    




(315770, 10)
(315770, 10)

In [5]:

    

print(np.amax(np.abs(pred - pred_original)))
plt.hist(pred - pred_original);
np.argmax(np.abs(pred - pred_original))
diff = np.abs(np.abs(pred - pred_original))
print('diff,',np.amax(diff))


    

    




0.0018973819
diff, 0.0018973819






    

In [48]:

    

diff = np.abs(pred - pred_original)
if (diff > 1.0).any() :
    print(np.where(diff))
    print('diff,',diff)


    

    




(array([     0,      0,      0, ..., 946040, 946040, 946040]), array([0, 1, 2, ..., 1, 2, 3]))
diff, [[1.9671917e-03 1.6583800e-03 1.3827783e-05 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]
 [2.9120028e-01 1.3768085e-05 1.5714478e-01 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]
 [4.0221941e-01 1.1783838e-02 1.4228137e-05 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]
 ...
 [3.4158826e-03 3.2285452e-03 0.0000000e+00 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]
 [7.3164701e-04 1.1019707e-03 1.3946974e-05 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]
 [8.3613396e-04 8.7004900e-04 1.3738375e-05 ... 0.0000000e+00
  0.0000000e+00 0.0000000e+00]]

In [45]:

    

diff


    

    
Out[45]:





array([[1.9671917e-03, 1.6583800e-03, 1.3827783e-05, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00],
       [2.9120028e-01, 1.3768085e-05, 1.5714478e-01, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00],
       [4.0221941e-01, 1.1783838e-02, 1.4228137e-05, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00],
       ...,
       [3.4158826e-03, 3.2285452e-03, 0.0000000e+00, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00],
       [7.3164701e-04, 1.1019707e-03, 1.3946974e-05, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00],
       [8.3613396e-04, 8.7004900e-04, 1.3738375e-05, ..., 0.0000000e+00,
        0.0000000e+00, 0.0000000e+00]], dtype=float32)

In [37]:

    

pred[946040, :4]


    

    
Out[37]:





array([9.9914742e-01, 1.0008526e+00, 4.9533377e-10, 4.9790455e-10],
      dtype=float32)

In [66]:

    




    

In [75]:

    

np.where(diff > 1.)


    

    
Out[75]:





(array([   188,   4437,   7850,  10111,  11478,  13693,  21808,  24959,
         27887,  30301,  32220,  32777,  34529,  38274,  42642,  44797,
         51224,  60785,  61890,  62827,  63162,  66036,  68741,  70308,
         72212,  73012,  73560,  74796,  74841,  75725,  83686,  86835,
         87631,  88612,  92408,  92705,  97529, 103487, 108790, 113241,
        124204, 126564, 127710, 130133, 133649, 146617, 148977, 151891,
        152319, 161569, 162825, 168739, 168751, 169882, 170746, 171070,
        171849, 172225, 172944, 175744, 176030, 180017, 182129, 182324,
        185003, 186570, 189042, 190570, 192993, 194930, 199685, 202512,
        202585, 205322, 207602, 209998, 214282, 216005, 219088, 221292,
        221466, 223513, 223555, 223665, 229158, 230142, 232466, 233951,
        234929, 236223, 253240, 254464, 259725, 261086, 265135, 269883,
        270270, 272269, 277860, 296335, 300207, 302662, 302873, 304270,
        304651, 311820, 319407, 322698, 326101, 326894, 327022, 330514,
        338857, 344539, 350488, 354730, 363839, 365901, 370501, 370780,
        373695, 374126, 380044, 382245, 383970, 384682, 386293, 386767,
        387167, 393395, 397487, 402940, 405552, 413067, 414541, 415576,
        418558, 419289, 419567, 434276, 435383, 438015, 445898, 448297,
        469741, 474310, 474945, 475499, 478171, 484927, 488284, 495754,
        496896, 498657, 503893, 512346, 513426, 515819, 526667, 541518,
        543333, 544383, 554161, 556997, 562688, 568581, 570876, 572342,
        574372, 580620, 580990, 581169, 581797, 584548, 585619, 591674,
        591926, 593955, 595573, 597409, 599417, 599692, 602493, 603059,
        612178, 614069, 618764, 619746, 619966, 620126, 622242, 623541,
        626018, 627496, 632846, 635936, 636429, 639197, 642545, 646627,
        652942, 662521, 664519, 667347, 672417, 673015, 677442, 687131,
        690053, 690981, 692346, 692980, 693352, 694396, 702797, 711024,
        711496, 717610, 719072, 722548, 738212, 738416, 738811, 741820,
        742773, 742779, 743903, 753311, 753404, 762293, 764010, 768542,
        768744, 771178, 772713, 775667, 776064, 776438, 779082, 784683,
        784720, 785226, 787054, 796250, 796289, 800596, 801317, 802246,
        803220, 804886, 805557, 806438, 807316, 809370, 813032, 814671,
        820806, 827178, 827306, 832139, 833720, 834465, 840381, 841314,
        841805, 846031, 851955, 855973, 859731, 861957, 862717, 865935,
        866069, 868090, 869334, 870919, 874340, 875466, 879972, 883295,
        886992, 887184, 891844, 893134, 897563, 900956, 901389, 902740,
        906288, 916885, 919146, 921545, 922417, 925620, 930973, 933118,
        939327, 939634, 943027, 944099, 945224]),
 array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2,
        0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]))

In [76]:

    

x = 722548
jets = 5

print(pred[x, :jets])
print(pred_original[x, :jets])
print(np.abs(pred_original[x, :jets] - pred[x,:jets]))


    

    




[2.0000000e+00 9.7428883e-09 2.2624598e-09 0.0000000e+00 0.0000000e+00]
[9.9996197e-01 4.3516338e-02 1.4662743e-05 0.0000000e+00 0.0000000e+00]
[1.0000380e+00 4.3516327e-02 1.4660480e-05 0.0000000e+00 0.0000000e+00]

In [ ]:

    

print(np.amax(np.abs(pred - pred_original)))
plt.hist(pred - pred_original);
np.argmax(np.abs(pred - pred_original))


    

In [21]:

    

C = (4.96392829e-07+ 1.40078935e-08+ 8.53559504e-06)
print(C)


    

    




9.0459957625e-06

In [15]:

    

x = [4.96392829e-07, 1.40078935e-08, 8.53559504e-06]


    

In [20]:

    

8.53559504e-06 * 221092.29901377592


    

    
Out[20]:





1.8871543308441827

In [23]:

    

a = (2 * 8.53559504e-06) /  9.0459957625e-06
print(a)


    

    




1.887154330844183

In [ ]:

    

np.abs(pred - pred_original)[2929, :]


    

In [ ]:

    

pred_original[2929, :]


    

In [ ]:

    

print(np.amax(np.abs(pred - pred_original)))
plt.hist(pred - pred_original);


    

In [ ]:

    

np.array_equal(pred_original,pred)


    

In [6]:

    

diff = pred - pred_original


    

In [7]:

    

diff.shape


    

    
Out[7]:





(944934, 10)

In [8]:

    

diff_scores = np.abs(diff)


    

In [16]:

    

diff_scores[0]


    

    
Out[16]:





array([1.9671917e-03, 1.6583800e-03, 1.3827783e-05, 1.3798038e-05,
       0.0000000e+00, 0.0000000e+00, 0.0000000e+00, 0.0000000e+00,
       0.0000000e+00, 0.0000000e+00], dtype=float32)

In [9]:

    

np.mean(diff_scores[0])


    

    
Out[9]:





0.09254789

In [ ]:

    

np.mean(diff_scores[0])


    

In [ ]:

    

np.mean(diff_scores[0])


    

In [ ]:

    

max_diff = np.amax(np.abs(diff))
# print(max_diff)
print("{:.6}".format(max_diff))
plt.hist(diff, bins=np.linspace(-max_diff, max_diff, 30))
# plt.hist(diff, bins=np.linspace(-1e-5, 1e-5, 30))
plt.yscale('log')
print(pred_original[np.abs(diff) > max_diff *0.9])


    

In [ ]:

    

a = np.array([[[9.4771385e-06],
  [8.9406967e-08],
  [0.0000000e+00],
  [4.3809414e-06],
  [2.3354590e-03],
  [2.3354590e-03],
  [2.3354590e-03],
  [2.3354590e-03],
  [2.3354404e-03],
  [2.3354404e-03]]]
)


    

In [ ]:

    

b = np.array([[[9.4838351e-06],
  [3.2189821e-08],
  [8.8993239e-09],
  [4.3864306e-06],
  [4.3864306e-06],
  [4.3864306e-06],
  [4.3864306e-06],
  [4.3864306e-06],
  [4.3864306e-06],
  [4.3864306e-06]]])


    

In [ ]:

    

np.amax(np.abs(a[0,0:4, :]- b[0, 0:4, :]))


    

In [ ]:

    

plt.hist((a[0, 0:4, :]- b[0, 0:4, :]).flatten())


    

In [ ]:

    

a.shape


    

In [ ]: