We will use distance between test segments computed in 140926-test-signal-jump to find sequence of segments that were likely together. Armed with this fact we can take the individual proababilities of each segment and combine it to form one probability that will be used to update the probabilites of all the segments in the sequence

the sequences are found using a greedy algoirthm that stops when a conflict is detected

the probabilities of segments should be combined by multiplying them, however this did not work well. Probably because the probabilites are not well calibrated. Taking the mean had a better effect.

Suppose you have a chain of segments: $i \in 1 \ldots N $

Each segment predicts a seizure $P_i$ or not $Q_i=1-P_i$

if a chain is negative then the probability is $\prod Q_i$ if a chain is positive the situation is more complex. There is a chance $U$ that a seizure detection even has happened and $V=1-U$ it did not. I estimate $U$ to be around $0.2$. So the probability is $\prod ( U * P_i + V*Q_i)$

or $\prod Q_i \times \Pi ( U \frac{P_i}{Q_i} +V )$

the ratio of positive to negative probability is just $r = \prod ( U \frac{P_i}{Q_i} +V )$ and probability is $1/(1+1/r)$



In [258]:

    
%matplotlib inline
from matplotlib import pylab as pl
import cPickle as pickle
import pandas as pd
import numpy as np
import os

individual segment probablility file



In [259]:

    
FNAME_IN = '/Users/udi/Downloads/github/SPC2014/submission.csv'

updated probability file



In [260]:

    
FNAME_OUT = '../submissions/141124-predict.18.csv'



In [261]:

    
!head {FNAME_IN}









    



clip,preictal
Dog_1_test_segment_0001.mat,0.523234
Dog_1_test_segment_0002.mat,0.415492
Dog_1_test_segment_0003.mat,0.410591
Dog_1_test_segment_0004.mat,0.572507
Dog_1_test_segment_0005.mat,0.502743
Dog_1_test_segment_0006.mat,0.484501
Dog_1_test_segment_0007.mat,0.414814
Dog_1_test_segment_0008.mat,0.500983
Dog_1_test_segment_0009.mat,0.397229



In [262]:

    
scores = pd.read_csv(FNAME_IN, index_col='clip', squeeze=True)
out_scores = scores.copy()



In [263]:

    
scores.hist()









    Out[263]:





<matplotlib.axes._subplots.AxesSubplot at 0x11e83f650>



In [264]:

    
scores['Dog_2_test_segment_0004.mat']









    Out[264]:





0.40474899999999997



In [265]:

    
targets = set(['_'.join(f.split('_')[:2]) for f in scores.index.values])
targets









    Out[265]:





{'Dog_1', 'Dog_2', 'Dog_3', 'Dog_4', 'Dog_5', 'Patient_1', 'Patient_2'}



In [266]:

    
W=0.65
T=0.2
D=-0.6

for target in targets:
    print
    d = np.load('/Users/udi/Downloads/kaggle/seizure-prediction/distance1/%s-test-jump-distance.npy'%target)
    N = d.shape[0]
    print target, N
    dord = np.unravel_index(d.ravel().argsort(),d.shape)
    Nsequences = N/6
    
    # find good pairs of segments that are likely to be paired in time
    next_segment = [-1]*N
    previous_segment = [-1]*N
    for i,(s1,s2) in enumerate(np.array(dord).T):
        dist = d[s1,s2]
        if dist > D: # from 140926-train-signal-jump
            break
        if next_segment[s1] != -1:
            #print i,'right conflict',dist
            continue
        if previous_segment[s2] != -1:
            #print i,'left conflict',dist
            continue
        c = 1
        j = s1
        while previous_segment[j] != -1:
            j = previous_segment[j]
            c += 1
        j = s2
        c += 1
        while next_segment[j] != -1:
            j = next_segment[j]
            c += 1
        if c > 6:
            continue
        next_segment[s1] = s2
        previous_segment[s2] = s1
#     if i < Nsequences:
#         print 'skip'
#         continue
    # check code
    for i in range(N):
        if next_segment[i] != -1:
            assert previous_segment[next_segment[i]] == i

    # find good sequences
    sequences = []
    for i in range(N):
        if previous_segment[i] == -1 and next_segment[i] != -1:
            j = i
            sequence = [j]
            while next_segment[j] != -1:
                j = next_segment[j]
                sequence.append(j)
            sequences.append(sequence)
    len_sequences = [len(sequence) for sequence in sequences]
    print '#sequences',len(sequences), '%segments that was sequenced',sum(len_sequences)/float(N), 'longest sequence', max(len_sequences)
    print sequences

    #compute probability for sequences
    sequences_prb = []
    for sequence in sequences:
        probs = np.array([scores['%s_test_segment_%04d.mat'%(target,s+1)] for s in sequence])
        wgts = np.exp(probs/T)
        wgts /= np.sum(wgts)
        p = np.dot(wgts, probs)
        sequences_prb.append(p)
    # fix probability for segments in sequences
    for p,sequence in zip(sequences_prb,sequences):
        # all segments in the same sequence will be assigned the same probability
        n = 1./len(sequence)
        for i, s in enumerate(sequence):
            out_scores['%s_test_segment_%04d.mat'%(target,s+1)] = W*scores['%s_test_segment_%04d.mat'%(target,s+1)] +(1.-W)*p









    



Dog_2 1000
#sequences 190 %segments that was sequenced 0.99 longest sequence 6
[[11, 336, 161, 793, 310], [13, 179, 925, 636, 444], [16, 862, 329, 618], [35, 175, 472, 152, 340, 606], [36, 480, 880, 304, 421, 569], [44, 346, 660, 383, 626, 21], [49, 902, 944, 490, 488, 337], [56, 559, 980, 287, 414, 464], [59, 234, 352, 820, 14, 42], [61, 297, 45, 504, 760, 630], [62, 859, 858, 785, 133, 30], [65, 755, 465, 459, 550, 871], [69, 68, 351, 830], [74, 302, 775, 588, 673, 184], [80, 776, 108, 293, 9, 47], [81, 141, 620, 573, 51, 930], [92, 306, 455, 101, 994, 331], [93, 598, 695, 699, 638, 33], [94, 973], [98, 223, 651], [99, 359, 326, 879, 813, 333], [104, 889, 798, 733, 50, 72], [106, 217, 250, 892, 702, 555], [109, 63, 404, 447, 365, 763], [114, 79, 87, 216, 469, 977], [115, 57, 979, 594, 185, 341], [129, 722, 116, 640], [146, 634, 281, 498, 535, 674], [148, 762, 978], [158, 198, 658, 652, 566, 113], [166, 105, 371, 197, 338], [173, 954, 761, 370, 608, 909], [176, 662, 897, 117, 987, 539], [193, 38, 821, 429, 887], [209, 509, 697, 339, 846, 937], [214, 419, 357, 221, 764, 252], [225, 384, 582, 251, 289, 616], [230, 402, 121, 267, 347, 46], [236, 783, 291, 548, 6, 386], [237, 857, 515, 969, 266, 532], [238, 578, 112, 400, 139], [249, 202, 600, 277, 196, 296], [256, 393, 2, 220, 482, 523], [258, 531, 201, 965, 835, 385], [260, 593, 22, 424, 312, 160], [270, 100, 507, 589, 602, 525], [284, 663, 716, 282, 826, 833], [294, 305, 283, 917, 136, 510], [295, 332, 466, 24, 558, 853], [299, 685, 831, 334, 298, 452], [303, 194, 877, 922, 984, 715], [313, 408, 707, 752], [319, 407, 819], [335, 467, 911, 103, 745, 561], [348, 240, 855, 836, 806, 503], [358, 164, 157, 731, 309, 723], [360, 413, 491], [363, 711, 328, 586, 991], [374, 130, 10, 572, 60, 495], [376, 215, 789, 948, 327, 438], [394, 703, 19, 330, 269], [403, 95, 131, 941, 955, 574], [406, 748, 318, 533, 89, 111], [415, 841, 832, 255, 612, 581], [416, 76, 502, 777], [417, 982, 300, 700, 665, 538], [420, 725, 369, 570, 924, 373], [422, 547, 366, 534, 477], [425, 643, 647], [432, 307, 445, 474, 471], [433, 200, 271, 617, 568, 468], [434, 720, 521, 993, 135], [436, 840, 856, 219, 536, 577], [437, 245, 546, 514, 657, 323], [440, 839], [449, 458, 661, 349], [450, 83, 920, 794, 751], [451, 58, 315, 772, 124, 91], [456, 874, 517, 943, 637, 957], [461, 423, 769, 321, 784], [475, 787], [484, 392], [486, 885, 805, 27, 378], [494, 147, 54, 850, 560, 929], [497, 900, 844, 738, 246, 625], [501, 278, 686, 718, 649, 726], [512, 401, 239, 122, 183, 967], [518, 659], [522, 809, 680], [528, 189, 460, 981, 155], [530, 778, 933], [537, 995, 77, 442], [551, 629, 31, 614, 361, 192], [553, 64, 435, 818, 165], [554, 153, 843], [563, 224, 162, 213, 5], [579, 144, 852, 986], [584, 549, 916, 285], [585, 261, 43, 754, 441, 687], [591, 903, 187, 48, 552, 542], [596, 996, 882, 782, 470], [601, 262, 653, 688, 747, 597], [603, 963, 757, 0], [610, 621], [628, 500, 997, 811, 485], [632, 623, 691, 908, 741, 156], [633, 273, 807, 150], [639, 205, 322, 233, 292], [644, 375, 190, 427, 481, 120], [646, 195], [648, 453, 431, 824, 82, 132], [650, 983], [654, 186, 487, 395, 272], [655, 631, 125, 712], [656, 627], [664, 232, 792], [666, 177, 127, 998, 170, 32], [667, 950, 443, 354], [671, 694, 67, 254, 350], [678, 635, 208, 286, 575, 682], [681, 41, 167, 896, 780], [689, 492, 975, 364, 211, 96], [693, 705, 590, 15, 571, 599], [706, 692, 556, 607], [708, 3, 226, 140, 838, 301], [709, 290, 849, 88, 947, 827], [710, 624, 97, 907, 142], [713, 342, 730, 397, 749, 834], [717, 968, 324, 675, 288], [719, 345, 454, 412, 765, 188], [721, 123, 280, 439, 817], [727, 25, 446, 520, 936, 229], [735, 390, 802, 398, 526], [736, 696, 381, 473, 949, 279], [739, 734, 970, 866, 85, 684], [742, 428, 23, 956, 935, 457], [743, 499, 479, 904, 609], [756, 753, 379, 567, 75, 53], [758, 462, 174, 163, 893, 516], [759, 84, 235, 118], [766, 714, 418, 815, 316, 851], [768, 396, 808], [771, 932, 972, 191, 138, 524], [786, 888, 128, 732, 906, 669], [788, 367, 86, 387, 641, 781], [790, 677, 26, 37, 276, 355], [795, 690, 344, 946, 847, 698], [797, 222, 990, 203, 275, 962], [799, 430, 910, 181, 263, 145], [803, 362, 562, 767, 508, 541], [804, 576, 182, 886, 28, 489], [812, 382, 529, 274, 248, 227], [814, 878, 988, 864, 527, 800], [822, 40, 611, 210, 399, 259], [823, 729, 343, 583, 107, 905], [828, 180, 20, 923, 228], [829, 895, 311, 178], [837, 953, 356], [842, 921, 770, 377, 672, 476], [845, 505, 746, 172, 724, 320], [848, 372, 592, 872, 206, 71], [861, 989, 513, 883, 587, 154], [867, 7, 595, 791, 773], [869, 264, 317, 939, 645, 478], [870, 212, 496, 992, 159], [873, 564, 959], [876, 448, 619, 810, 18, 974], [881, 668, 483], [884, 898, 137, 615, 353, 410], [891, 171, 391, 890, 580, 927], [894, 12, 875, 999, 90, 265], [899, 368, 110, 779, 389, 241], [901, 701, 8, 914, 642, 126], [912, 151, 825, 543], [913, 557, 169, 257, 73, 380], [915, 134, 854, 934, 231, 704], [918, 966, 39, 463, 960, 613], [919, 52, 737, 268, 728], [926, 961, 4, 774, 816, 868], [928, 409, 218, 622, 796, 744], [931, 34, 143, 605, 253, 405], [940, 244, 670, 740, 519], [942, 683], [945, 676, 243, 411, 750, 511], [951, 493, 1, 565, 976, 314], [952, 540, 679, 119, 325, 545], [958, 506, 199, 426, 168], [964, 78, 102, 865, 604, 247], [971, 863], [985, 29, 801, 66, 938, 17]]

Dog_3 907
#sequences 175 %segments that was sequenced 0.97243660419 longest sequence 6
[[3, 81, 539, 270, 702, 582], [13, 226, 111, 33, 645], [14, 116, 537, 654], [22, 211, 854, 889, 176, 486], [32, 396, 59, 569, 594, 296], [36, 714, 218, 630], [44, 456, 906, 820, 521, 538], [45, 453, 480, 215, 688, 150], [48, 728, 665, 363, 802, 762], [63, 432, 669, 664, 739, 483], [67, 708, 241, 715], [68, 522, 598, 37, 618, 54], [70, 368, 672, 600, 96], [76, 450, 704, 120], [78, 513, 187, 56, 302, 766], [83, 170], [85, 497, 308, 718, 5, 751], [87, 847, 661, 720, 519, 327], [88, 16, 238, 581], [95, 19, 818, 458, 694, 528], [97, 49, 850], [101, 230], [107, 851, 246, 683, 596, 485], [118, 72, 635, 481, 874, 89], [121, 34, 860, 679, 17, 684], [131, 599, 39, 434, 367, 52], [135, 603, 852, 873, 780, 243], [137, 656, 564, 696], [151, 615, 415, 400, 423, 748], [152, 40, 154, 839, 98, 756], [159, 210, 648, 698, 534, 791], [162, 863, 202, 307, 546, 742], [164, 427, 279, 160], [166, 901, 468, 487, 245, 342], [167, 244, 161], [169, 779, 42, 588, 53, 275], [172, 355], [174, 82, 283, 188, 310, 811], [184, 35, 445, 580, 366, 51], [192, 631, 551, 12, 345, 786], [193, 145, 837], [197, 832, 626, 104, 325, 277], [199, 147, 2, 171, 182, 347], [201, 372, 565, 893, 710, 828], [209, 177, 770, 776, 231], [224, 848, 727, 690, 55, 821], [227, 808, 524, 11, 760, 701], [228, 394, 280, 31, 709, 282], [235, 625, 540, 252, 619, 735], [251, 321], [255, 774, 681, 234, 767, 21], [256, 122, 510, 130, 27, 50], [259, 102, 526, 764, 382, 872], [263, 389, 498, 110, 117, 207], [264, 433, 844, 757, 262, 655], [267, 650, 562, 299, 712, 827], [271, 298, 38, 428, 388, 426], [273, 508, 721, 425, 717, 379], [286, 555, 680, 525, 41], [289, 608, 689, 148, 447, 272], [293, 733, 198, 297, 418, 319], [300, 612, 305, 23], [306, 897, 361, 265], [309, 876, 375, 344, 94, 132], [313, 499, 18, 667, 755, 758], [315, 100], [324, 179, 303, 724, 489], [329, 572, 217, 268, 203, 511], [331, 191, 260, 501, 369, 552], [334, 730], [338, 205, 559, 406, 333, 804], [339, 20, 143, 459, 846, 311], [343, 896, 741, 402, 640, 216], [346, 576, 587, 662], [350, 891, 129, 750, 529], [351, 149, 649], [354, 413, 138, 348, 168, 359], [358, 92, 431, 890, 105, 813], [362, 841, 798, 561, 810, 189], [365, 532, 461, 291], [371, 376, 185, 250, 886, 77], [373, 606, 208, 787, 782, 71], [374, 115, 341, 639, 793], [378, 644, 797], [380, 417], [386, 557, 765, 274, 647, 287], [393, 407, 146, 693], [397, 505, 318, 337], [409, 805, 421, 530, 10, 660], [411, 697, 899, 183, 340, 642], [412, 575, 290, 531, 401, 477], [440, 25, 732, 212, 866, 336], [443, 479, 200, 678, 233, 180], [444, 220, 91, 429, 157, 294], [454, 119], [457, 106], [462, 370, 527, 584], [466, 424, 139, 194, 549], [473, 254], [474, 822, 835], [478, 883, 258, 550, 703, 57], [490, 853, 577], [492, 834, 28, 613, 836, 422], [500, 682, 628, 123], [504, 620, 126, 882, 384], [509, 74, 610, 520, 623, 391], [512, 436, 403, 633, 186, 364], [514, 24], [517, 206, 114, 452, 794, 439], [535, 892, 885, 545, 591], [542, 178, 0, 796, 304], [543, 190, 229, 590, 112, 410], [544, 536, 506], [556, 257, 460, 488, 46, 395], [558, 240], [567, 455, 643, 515, 141, 566], [579, 113, 784, 441, 249], [585, 223, 125, 496], [602, 745, 288, 595, 142], [604, 140, 795, 548, 586, 62], [614, 716, 158, 93, 888, 824], [616, 759, 204, 833, 442, 8], [617, 632, 788, 43, 624, 484], [622, 387, 695, 840, 686], [627, 773, 266, 589, 856, 747], [634, 219, 295, 335, 446, 181], [637, 641, 349, 451, 322, 247], [653, 7, 420, 173, 73, 865], [657, 467, 66], [658, 878, 475, 404, 578, 278], [666, 153, 605], [668, 700, 502, 312, 905, 214], [670, 323], [673, 465, 109, 242, 838, 849], [685, 213, 175, 675, 674, 377], [691, 875, 330, 405, 124], [705, 692, 533, 4, 734, 292], [706, 570, 470], [719, 144, 740, 746, 884, 90], [722, 713, 86, 155], [729, 269, 541, 807, 864, 790], [731, 469, 871], [736, 597, 601, 495, 753, 285], [738, 769, 103, 136, 659, 6], [743, 30, 563, 237, 408, 754], [744, 777, 99, 638, 471, 284], [749, 857, 571, 503, 895, 568], [752, 464, 789, 414, 353, 677], [761, 523], [771, 768, 232, 225], [772, 163, 26], [778, 60, 281, 554, 801, 221], [781, 507, 493, 763, 1, 573], [785, 356, 726], [799, 862, 806, 845, 646], [800, 398, 317, 320, 699], [803, 817, 687, 449, 261, 195], [814, 47, 870, 609, 58, 887], [815, 357, 707, 636, 583], [816, 792, 783, 435, 61, 352], [819, 843, 671, 239, 128, 29], [825, 903, 861, 156], [826, 314, 64, 775], [829, 472, 399, 236, 812, 416], [830, 607, 711, 165, 127, 560], [831, 9, 842, 438], [855, 84, 430, 463, 652, 69], [858, 360, 65], [859, 494, 276, 79, 381, 332], [867, 737, 880, 133, 676, 809], [877, 547, 316, 621, 491, 725], [879, 80, 553, 196, 437, 894], [900, 385, 392], [902, 108, 482, 723, 593, 419], [904, 448]]

Dog_1 502
#sequences 97 %segments that was sequenced 0.994023904382 longest sequence 6
[[3, 493, 475, 178], [6, 73, 384, 444, 216, 155], [10, 177, 271, 255, 60], [13, 231, 159], [14, 317, 276, 15, 123], [16, 453, 425, 285, 191], [21, 32, 236, 199, 218], [22, 291, 381, 92, 268, 86], [23, 359, 409, 491], [28, 124, 88, 215, 197, 242], [29, 398, 12, 224, 31, 0], [33, 390, 383, 309, 119, 396], [40, 256, 262, 361, 428], [41, 357, 339, 153, 174, 364], [46, 464, 135, 89, 264, 368], [47, 211, 117, 479, 185, 246], [53, 126, 401, 250, 9, 172], [68, 406, 282, 161, 468, 427], [71, 382, 192, 190, 463, 328], [72, 323, 351, 496, 385, 275], [74, 290], [76, 127, 146, 106, 389, 349], [77, 204, 347, 214, 380, 462], [78, 375, 227, 109, 202, 405], [79, 48, 244, 114], [81, 418, 208, 239, 67, 345], [93, 321, 186, 473, 210, 298], [95, 452, 407, 445, 115, 358], [103, 171, 252, 175, 412, 154], [105, 205, 187, 207, 82, 472], [110, 85, 7, 258], [128, 373, 477, 356, 50, 500], [133, 288, 260, 490, 331, 424], [137, 162, 287, 129, 344], [139, 18, 313, 65], [149, 116, 97, 107, 219, 284], [150, 148, 112, 253, 283], [152, 179, 194, 377, 99, 87], [158, 140, 411, 143, 352, 474], [160, 434, 267, 293, 442, 221], [169, 302, 343, 232], [184, 173, 485, 96, 156, 480], [225, 397, 228, 448, 35, 318], [226, 66, 134, 476, 44, 108], [234, 329, 438, 147, 136], [235, 37, 84, 336, 295], [237, 426, 118], [238, 492, 58, 371, 222], [247, 223, 249, 102, 62, 42], [251, 266, 483, 125, 61], [257, 122, 326, 432, 332], [261, 248, 416, 132, 70, 495], [265, 144], [269, 327, 270, 52, 8, 98], [274, 100], [278, 36, 198, 387, 69, 11], [289, 470, 330], [294, 341, 176, 245, 142, 292], [299, 478, 90, 353, 272], [301, 366, 342, 30], [303, 19], [304, 230, 410, 471, 286, 201], [306, 421, 1], [311, 212, 56, 24, 138], [315, 163, 131, 80], [320, 57, 413, 75, 196, 27], [340, 281, 229, 333, 217, 233], [348, 213, 360, 441, 4, 220], [354, 181, 431, 338, 200], [355, 83, 39], [365, 277, 322, 130, 2, 378], [367, 484, 180, 335, 296, 486], [369, 439, 488, 362, 312, 422], [370, 316, 300, 120, 141, 400], [374, 408, 447, 104, 297, 429], [379, 443, 449, 38, 399, 417], [388, 113, 280, 437, 168], [391, 94, 337], [393, 145, 350, 319, 324, 334], [414, 459, 45, 363, 55], [415, 273], [419, 195, 5, 420], [423, 465, 151, 386, 64], [450, 395, 455, 34, 307, 376], [451, 17, 454, 314, 254, 433], [457, 460, 310, 456, 372], [461, 193, 91, 167, 188, 259], [466, 209, 263, 346, 26, 170], [467, 243, 101, 54, 240, 394], [469, 51, 497, 43, 20], [481, 489, 430, 308, 121, 404], [482, 241, 440, 49, 111, 59], [487, 206, 25, 157, 435, 166], [494, 63, 189, 458], [498, 305, 403], [499, 446, 182, 392, 164], [501, 325, 436, 165, 279]]

Dog_4 990
#sequences 171 %segments that was sequenced 0.990909090909 longest sequence 6
[[0, 399, 297, 940, 455, 284], [1, 260, 101, 478, 415, 236], [7, 759, 635, 264, 216, 848], [12, 184, 479, 829, 139, 724], [14, 581, 429, 714, 660, 582], [15, 24, 960, 519, 931, 965], [25, 315, 527, 323, 595, 133], [31, 328, 740, 637, 167, 942], [33, 631, 493, 561, 421, 74], [34, 38, 978, 678], [36, 964, 75, 141, 790, 775], [37, 340, 532, 632, 883, 518], [40, 387, 810, 690, 930, 679], [44, 875, 116, 405, 118, 565], [46, 402, 971, 446, 506, 901], [55, 170, 319, 767, 925, 252], [56, 431, 566, 815, 124, 296], [60, 127, 847, 634, 778, 285], [63, 528, 362, 359, 346, 332], [67, 937, 987, 148, 948, 72], [85, 959, 897, 388, 962, 580], [87, 584, 902, 99, 289, 50], [92, 643, 212, 240, 335, 501], [93, 80], [103, 737, 721, 709, 131, 440], [112, 827, 155, 106, 890, 244], [120, 122, 281, 727, 406, 820], [123, 383, 809, 970, 82, 333], [125, 647, 374, 624, 312, 730], [128, 134], [136, 853, 417, 230, 494, 966], [138, 235, 613, 208, 349, 413], [140, 499, 604, 696, 379, 720], [142, 341, 860, 409, 655, 791], [159, 628, 261, 921, 375, 586], [163, 817, 648, 540, 39, 626], [172, 680, 295, 68, 214, 194], [175, 976, 16, 688, 924, 262], [176, 743, 196, 972, 702, 173], [177, 795, 520, 713, 439, 433], [197, 545, 531, 17, 756, 674], [204, 749, 876, 471, 114, 866], [211, 638, 376, 102, 373, 259], [220, 502, 770, 908, 985, 828], [225, 813, 344, 907, 465, 652], [242, 881, 207, 878, 253, 956], [256, 516, 473, 983, 597, 485], [263, 288, 186, 22, 444, 859], [276, 291, 313, 662, 272, 305], [279, 222, 327, 448, 729, 504], [282, 957, 497, 269, 656, 426], [294, 156, 357, 733, 698, 891], [299, 152, 183, 456, 861, 880], [300, 239, 926, 851, 392, 474], [303, 11, 10, 191, 89, 364], [309, 361, 187, 585, 728, 154], [310, 377, 783, 830, 672, 412], [314, 974, 542, 741, 832, 164], [320, 507, 824, 895, 715, 378], [329, 685, 523, 673, 686, 543], [330, 98, 144, 747, 988, 554], [338, 768, 703, 927, 511, 893], [343, 951, 943, 203, 760, 608], [350, 766, 835, 807, 837, 365], [353, 107, 664, 546, 18, 571], [372, 165, 923, 961, 66, 464], [389, 355, 337, 805, 583, 569], [390, 578, 599, 706, 910, 742], [394, 51, 564, 739, 316, 487], [395, 110, 419, 606, 977, 217], [396, 984, 257, 658, 954, 663], [403, 162, 659, 754, 798, 53], [404, 846, 58, 318, 229, 731], [411, 401, 286, 514, 398, 657], [424, 900, 251, 618, 232, 691], [425, 722, 258], [430, 894, 711, 23, 45, 462], [434, 27, 748, 787, 290, 933], [436, 734, 842, 348, 562, 416], [441, 619, 121, 573, 324, 544], [443, 794, 843, 77, 234, 808], [451, 449, 935, 873, 19, 592], [458, 687, 950, 629, 917, 854], [461, 906, 489, 322, 292, 552], [469, 796, 146, 226, 147, 200], [472, 958, 834, 761, 968, 109], [481, 877, 826, 973, 969, 247], [490, 919, 354, 90, 625, 548], [492, 188, 2, 505, 306, 3], [496, 273, 932, 73, 418], [498, 755, 718, 453, 168, 223], [500, 301, 233, 231, 459, 612], [509, 423, 13, 321, 132, 889], [513, 630, 366, 454, 428, 438], [515, 547, 605, 553, 52, 683], [522, 182, 29], [524, 850, 670, 797, 525, 945], [535, 278, 475, 677, 205, 601], [536, 836, 869, 115, 503, 166], [537, 248, 572, 905, 886, 477], [551, 28, 8, 589, 725, 801], [555, 920, 967, 654, 468, 470], [558, 484, 266, 979, 181, 852], [559, 903, 675, 936, 97, 145], [570, 385, 726, 526, 250], [575, 151, 71, 574, 283, 717], [577, 81, 653, 550, 704, 482], [588, 530, 407, 339, 457, 363], [596, 694, 735, 35, 466, 275], [598, 447, 645], [602, 265, 255, 224, 47, 195], [610, 667, 594, 342, 360, 169], [614, 137, 135, 311, 157, 111], [615, 369, 228, 892, 784, 368], [616, 517, 408, 215, 849, 280], [620, 178, 567, 161, 896, 76], [623, 975, 627, 593, 331, 947], [636, 867, 946, 707, 84, 367], [641, 65, 888, 644, 757, 336], [642, 96, 665, 414, 270, 317], [651, 668, 736, 833, 510, 149], [661, 61], [684, 914, 79, 126, 249, 119], [692, 83, 245, 695, 746, 541], [697, 871, 100], [699, 622, 666, 307, 753, 862], [710, 386, 708, 911, 856, 792], [712, 671, 210], [719, 928, 435, 922, 237, 804], [723, 206, 639, 752, 776, 20], [738, 48, 70, 483, 840, 986], [745, 351, 391, 4, 774], [750, 982, 352, 744, 108, 590], [751, 287, 274, 5, 370, 432], [762, 865, 870, 302, 189, 855], [763, 193, 885, 676, 649, 32], [764, 86, 789, 213, 872, 913], [765, 646, 650, 158, 427, 371], [771, 944, 909, 380, 160, 788], [777, 912, 486, 825, 129, 915], [779, 732, 64, 603, 591, 803], [786, 533, 410, 939, 617, 839], [806, 780, 549, 609, 293, 452], [811, 938, 347, 450, 221, 521], [812, 326, 219, 989, 62, 420], [814, 802, 793, 700, 772, 201], [816, 400, 246, 277, 192, 334], [818, 304, 397, 874, 693, 238], [821, 6, 104, 105, 382], [822, 781, 705, 467, 701], [823, 480, 180, 199, 57, 841], [831, 463, 539, 952, 785, 91], [838, 576, 579, 916, 568, 9], [844, 393, 113, 78, 356, 782], [845, 981, 254, 884, 381, 198], [858, 202], [863, 460, 21, 799, 179, 495], [864, 512, 43, 882, 267, 769], [868, 185, 30, 560, 209, 143], [879, 227, 358, 773, 669, 689], [887, 819, 563, 26, 243, 325], [898, 445, 633], [899, 556, 174, 508, 488, 529], [904, 963, 117, 587, 600, 69], [918, 268, 442, 934, 171, 241], [929, 476, 534, 218, 94, 59], [941, 758, 345, 491, 800, 681], [949, 150, 153, 88], [953, 42, 298, 857, 607, 41], [955, 621, 611, 190, 682, 49], [980, 422, 716, 130]]

Dog_5 191
#sequences 31 %segments that was sequenced 0.900523560209 longest sequence 6
[[1, 131, 105, 139, 147, 154], [3, 59, 30, 121, 12, 184], [11, 100, 79, 119, 55, 2], [17, 155, 28, 125, 98, 185], [29, 122, 71, 35, 22, 112], [34, 49, 27, 25, 165, 110], [38, 84, 107], [41, 21], [50, 179, 187, 57, 133, 126], [56, 170, 13, 58, 144, 14], [60, 80, 96, 15, 152, 5], [65, 111, 8], [67, 108, 52, 90, 159], [86, 82, 157, 24, 32, 74], [87, 183, 77, 113, 114, 135], [94, 161, 171, 46, 176, 39], [99, 43, 129, 75, 178, 88], [104, 162, 118, 66, 97, 76], [124, 93, 148, 136, 169, 132], [128, 48, 42, 160, 0, 73], [134, 85, 78, 158, 102, 189], [138, 92, 109, 44, 10, 7], [141, 95, 91, 115, 117, 163], [145, 130, 103, 123, 120, 180], [146, 149, 6, 36, 140], [153, 40, 150, 164, 142, 101], [166, 182, 33, 23, 190, 106], [167, 127, 69, 188], [168, 173, 151, 18, 63, 137], [175, 81, 172, 186, 83, 20], [177, 26, 143, 68, 62, 51]]

Patient_2 150
#sequences 28 %segments that was sequenced 0.96 longest sequence 6
[[1, 84, 105, 134, 3, 138], [4, 0], [6, 145, 106, 33], [8, 63, 122], [12, 98, 50, 144, 85, 133], [21, 56, 16, 43, 27, 112], [25, 76], [29, 71, 75, 58, 10], [30, 49, 135, 57, 28, 24], [37, 14, 11, 36, 109, 136], [39, 80, 67, 86, 96, 119], [42, 47, 140, 72, 93, 38], [46, 130, 34, 40, 2, 22], [51, 44, 129, 115, 55], [60, 141, 7, 64, 95], [62, 31, 121, 137, 18], [73, 53, 20], [78, 91, 26, 100, 125, 61], [90, 48, 35], [92, 23, 68, 126, 101, 32], [94, 41, 79, 9, 17, 52], [108, 124, 77, 139, 45, 107], [110, 65, 69, 87, 88], [113, 15, 123, 5, 103, 66], [114, 81, 82, 131, 97, 59], [127, 102, 99, 128, 118, 149], [142, 147, 146, 74, 13, 54], [148, 19, 89, 117, 120, 116]]

Patient_1 195
#sequences 48 %segments that was sequenced 0.830769230769 longest sequence 6
[[3, 52], [8, 14, 27, 95, 5, 18], [11, 40, 53, 28], [16, 30, 22, 74, 71, 81], [17, 63], [21, 64], [23, 87, 36, 61], [24, 113, 191, 137, 47, 99], [31, 15, 20], [33, 72, 80], [37, 48, 42], [46, 32, 44, 65, 123, 139], [54, 19, 13], [55, 127], [57, 56, 66, 68, 51], [58, 128, 185], [60, 92, 101, 83], [67, 141, 182, 135], [73, 102, 96], [78, 77, 35, 109, 157], [86, 45], [89, 110, 146, 170, 118], [90, 147], [93, 153], [107, 115], [108, 88], [116, 155], [122, 91, 160], [129, 84], [130, 114, 117], [131, 151, 103], [132, 165], [138, 158, 97, 179, 112], [149, 111], [150, 193, 189, 181, 187, 177], [152, 190, 145, 162, 100], [154, 142], [156, 183, 104, 136, 167], [166, 148, 159], [168, 164], [169, 134, 192, 121], [171, 119, 94, 43], [173, 124, 186, 106], [174, 125], [176, 163], [180, 50, 25], [184, 82, 79, 105], [194, 140, 39]]



In [267]:

    
out_scores = out_scores/out_scores.max()



In [268]:

    
out_scores.to_csv(FNAME_OUT, header=True)



In [269]:

    
!paste {FNAME_IN} {FNAME_OUT} | head









    



clip,preictal	clip,preictal
Dog_1_test_segment_0001.mat,0.523234	Dog_1_test_segment_0001.mat,0.5875265952652561
Dog_1_test_segment_0002.mat,0.415492	Dog_1_test_segment_0002.mat,0.4458150441395031
Dog_1_test_segment_0003.mat,0.410591	Dog_1_test_segment_0003.mat,0.430835965609518
Dog_1_test_segment_0004.mat,0.572507	Dog_1_test_segment_0004.mat,0.5607354288829067
Dog_1_test_segment_0005.mat,0.502743	Dog_1_test_segment_0005.mat,0.5039016486680596
Dog_1_test_segment_0006.mat,0.484501	Dog_1_test_segment_0006.mat,0.4950969930929728
Dog_1_test_segment_0007.mat,0.414814	Dog_1_test_segment_0007.mat,0.43469427030119906
Dog_1_test_segment_0008.mat,0.500983	Dog_1_test_segment_0008.mat,0.5044222798421228
Dog_1_test_segment_0009.mat,0.397229	Dog_1_test_segment_0009.mat,0.4092499631569485



In [270]:

    
out_scores['Dog_2_test_segment_0004.mat']









    Out[270]:





0.41551970377309527



In [271]:

    
df = pd.DataFrame()
df['in'] = pd.read_csv(FNAME_IN, index_col='clip', squeeze=True) #64
df['out'] = pd.read_csv(FNAME_OUT, index_col='clip', squeeze=True)
df['best'] = pd.read_csv('../submissions/141116-predict.8.csv', index_col='clip', squeeze=True)



In [272]:

    
pd.scatter_matrix(df,figsize=(6, 6), diagonal='kde');



In [272]: