Comments (7)
kk7nc
commented on August 16, 2024
Thank you for your question,
The RMDl contains many parameters as we explain in the main page,
but in your case can you run it for 9 models and set plot=True, and send me your plot?
It is needed to run for all models then it generates the plot,
Also for simplisity, we set Glove with 50 dimensions, but the results are reported by Glove 300 dimensions,
The results are based on the random model, but it's not huge differences, maybe 2-3 percents higher or lower.
from rmdl.
kk7nc
commented on August 16, 2024
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
RMDL: Random Multimodel Deep Learning for Classification
* Copyright (C) 2018 Kamran Kowsari <[email protected]>
* Last Update: Oct 26, 2018
* This file is part of RMDL project, University of Virginia.
* Free to use, change, share and distribute source code of RMDL
* Refrenced paper : RMDL: Random Multimodel Deep Learning for Classification
* Link: https://dl.acm.org/citation.cfm?id=3206111
* Refrenced paper : An Improvement of Data Classification using Random Multimodel Deep Learning (RMDL)
* Link : http://www.ijmlc.org/index.php?m=content&c=index&a=show&catid=79&id=823
* Comments and Error: email: [email protected]
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
import os
from RMDL import text_feature_extraction as txt
from sklearn.model_selection import train_test_split
from RMDL.Download import Download_WOS as WOS
import numpy as np
from RMDL import RMDL_Text as RMDL
if __name__ == "__main__":
path_WOS = WOS.download_and_extract()
fname = os.path.join(path_WOS,"WebOfScience/WOS11967/X.txt")
fnamek = os.path.join(path_WOS,"WebOfScience/WOS11967/Y.txt")
with open(fname, encoding="utf-8") as f:
content = f.readlines()
content = [txt.text_cleaner(x) for x in content]
with open(fnamek) as fk:
contentk = fk.readlines()
contentk = [x.strip() for x in contentk]
Label = np.matrix(contentk, dtype=int)
Label = np.transpose(Label)
np.random.seed(7)
print(Label.shape)
X_train, X_test, y_train, y_test = train_test_split(content, Label, test_size=0.2, random_state=4)
batch_size = 100
sparse_categorical = True
n_epochs = [5000, 500, 500] ## DNN--RNN-CNN
Random_Deep = [0, 0, 3] ## DNN--RNN-CNN
RMDL.Text_Classification(X_train, y_train, X_test, y_test,
batch_size=batch_size,
plot=True,
sparse_categorical=sparse_categorical,
random_deep=Random_Deep,
random_optimizor=False,
GloVe_file="glove.6B.50d.txt",
EMBEDDING_DIM = 50,
dropout=0.1,
epochs=n_epochs)
from rmdl.
zhuzhangli
commented on August 16, 2024
@kk7nc Thank you kk
I will try it first, then show the plot results to you.
from rmdl.
zhuzhangli
commented on August 16, 2024
I am very sorry that I did not output plot for the result of the run this time.
But could you take a look at the problem for me first? Here is the output from my run:
Using TensorFlow backend.
[nltk_data] Downloading package stopwords to /home/lc-lzt/nltk_data...
[nltk_data] Package stopwords is already up-to-date!
sys.version_info(major=3, minor=6, micro=5, releaselevel='final', serial=0)
sys.version_info(major=3, minor=6, micro=5, releaselevel='final', serial=0)
(11967, 1)
Done1
tf-idf with 51346 features
Found 59409 unique tokens.
(11967, 500)
Total 400000 word vectors.
33
DNN 0
<keras.optimizers.Adagrad object at 0x7ff100800e80>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 76s - loss: 1.5623 - acc: 0.5549 - val_loss: 0.7445 - val_acc: 0.7878
Epoch 00001: val_acc improved from -inf to 0.78780, saving model to weights\weights_DNN_0.hdf5
Epoch 2/100
- 19s - loss: 0.3373 - acc: 0.9161 - val_loss: 0.6406 - val_acc: 0.8074
Epoch 00002: val_acc improved from 0.78780 to 0.80744, saving model to weights\weights_DNN_0.hdf5
Epoch 3/100
- 25s - loss: 0.1039 - acc: 0.9823 - val_loss: 0.6343 - val_acc: 0.8175
Epoch 00003: val_acc improved from 0.80744 to 0.81746, saving model to weights\weights_DNN_0.hdf5
Epoch 4/100
- 8s - loss: 0.0433 - acc: 0.9942 - val_loss: 0.6449 - val_acc: 0.8158
Epoch 00004: val_acc did not improve from 0.81746
Epoch 5/100
- 19s - loss: 0.0244 - acc: 0.9979 - val_loss: 0.6522 - val_acc: 0.8150
Epoch 00005: val_acc did not improve from 0.81746
Epoch 6/100
- 19s - loss: 0.0152 - acc: 0.9993 - val_loss: 0.6582 - val_acc: 0.8150
Epoch 00006: val_acc did not improve from 0.81746
Epoch 7/100
- 8s - loss: 0.0107 - acc: 0.9996 - val_loss: 0.6668 - val_acc: 0.8187
Epoch 00007: val_acc improved from 0.81746 to 0.81871, saving model to weights\weights_DNN_0.hdf5
Epoch 8/100
- 12s - loss: 0.0079 - acc: 0.9997 - val_loss: 0.6742 - val_acc: 0.8162
Epoch 00008: val_acc did not improve from 0.81871
Epoch 9/100
- 9s - loss: 0.0060 - acc: 0.9999 - val_loss: 0.6830 - val_acc: 0.8150
...
Epoch 00098: val_acc did not improve from 0.81871
Epoch 99/100
- 34s - loss: 8.9374e-05 - acc: 1.0000 - val_loss: 0.8602 - val_acc: 0.8141
Epoch 00099: val_acc did not improve from 0.81871
Epoch 100/100
- 42s - loss: 1.0079e-04 - acc: 1.0000 - val_loss: 0.8608 - val_acc: 0.8141
Epoch 00100: val_acc did not improve from 0.81871
DNN 1
<keras.optimizers.RMSprop object at 0x7fefbe6ed588>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 17s - loss: 3.1534 - acc: 0.0760 - val_loss: 2.6202 - val_acc: 0.1541
Epoch 00001: val_acc improved from -inf to 0.15414, saving model to weights\weights_DNN_1.hdf5
Epoch 2/100
- 33s - loss: 2.3482 - acc: 0.2238 - val_loss: 1.9169 - val_acc: 0.3454
Epoch 00002: val_acc improved from 0.15414 to 0.34545, saving model to weights\weights_DNN_1.hdf5
Epoch 3/100
- 41s - loss: 1.6699 - acc: 0.4090 - val_loss: 1.4863 - val_acc: 0.5017
Epoch 00003: val_acc improved from 0.34545 to 0.50167, saving model to weights\weights_DNN_1.hdf5
Epoch 4/100
- 33s - loss: 1.2221 - acc: 0.5544 - val_loss: 1.3064 - val_acc: 0.6170
Epoch 00004: val_acc improved from 0.50167 to 0.61696, saving model to weights\weights_DNN_1.hdf5
Epoch 5/100
- 28s - loss: 0.9206 - acc: 0.6634 - val_loss: 1.3379 - val_acc: 0.6232
Epoch 00005: val_acc improved from 0.61696 to 0.62322, saving model to weights\weights_DNN_1.hdf5
Epoch 6/100
- 39s - loss: 0.7175 - acc: 0.7451 - val_loss: 1.4184 - val_acc: 0.6763
Epoch 00006: val_acc improved from 0.62322 to 0.67627, saving model to weights\weights_DNN_1.hdf5
Epoch 7/100
- 35s - loss: 0.5686 - acc: 0.8055 - val_loss: 1.4307 - val_acc: 0.6951
Epoch 00007: val_acc improved from 0.67627 to 0.69507, saving model to weights\weights_DNN_1.hdf5
Epoch 8/100
- 23s - loss: 0.4696 - acc: 0.8438 - val_loss: 1.4596 - val_acc: 0.7114
Epoch 00008: val_acc improved from 0.69507 to 0.71136, saving model to weights\weights_DNN_1.hdf5
Epoch 9/100
- 17s - loss: 0.3915 - acc: 0.8760 - val_loss: 1.5512 - val_acc: 0.7068
Epoch 00009: val_acc did not improve from 0.71136
Epoch 10/100
- 55s - loss: 0.3325 - acc: 0.8941 - val_loss: 1.6379 - val_acc: 0.7197
Epoch 00010: val_acc improved from 0.71136 to 0.71972, saving model to weights\weights_DNN_1.hdf5
Epoch 11/100
- 35s - loss: 0.2691 - acc: 0.9149 - val_loss: 1.7950 - val_acc: 0.7135
Epoch 00011: val_acc did not improve from 0.71972
Epoch 12/100
- 36s - loss: 0.2382 - acc: 0.9272 - val_loss: 1.7445 - val_acc: 0.7297
Epoch 00012: val_acc improved from 0.71972 to 0.72974, saving model to weights\weights_DNN_1.hdf5
Epoch 13/100
- 32s - loss: 0.2099 - acc: 0.9371 - val_loss: 1.9434 - val_acc: 0.7318
Epoch 00013: val_acc improved from 0.72974 to 0.73183, saving model to weights\weights_DNN_1.hdf5
Epoch 14/100
- 7s - loss: 0.1833 - acc: 0.9481 - val_loss: 1.9385 - val_acc: 0.7176
Epoch 00014: val_acc did not improve from 0.73183
Epoch 15/100
- 63s - loss: 0.1620 - acc: 0.9525 - val_loss: 1.9940 - val_acc: 0.7314
Epoch 00015: val_acc did not improve from 0.73183
Epoch 16/100
- 57s - loss: 0.1546 - acc: 0.9596 - val_loss: 2.0250 - val_acc: 0.7226
Epoch 00016: val_acc did not improve from 0.73183
Epoch 17/100
- 62s - loss: 0.1299 - acc: 0.9686 - val_loss: 2.0776 - val_acc: 0.7393
Epoch 00017: val_acc improved from 0.73183 to 0.73935, saving model to weights\weights_DNN_1.hdf5
Epoch 18/100
- 17s - loss: 0.1250 - acc: 0.9650 - val_loss: 2.1323 - val_acc: 0.7406
Epoch 00018: val_acc improved from 0.73935 to 0.74060, saving model to weights\weights_DNN_1.hdf5
Epoch 19/100
- 8s - loss: 0.1254 - acc: 0.9701 - val_loss: 2.1342 - val_acc: 0.7398
Epoch 00019: val_acc did not improve from 0.74060
Epoch 20/100
- 25s - loss: 0.0996 - acc: 0.9753 - val_loss: 2.1777 - val_acc: 0.7419
Epoch 00020: val_acc improved from 0.74060 to 0.74185, saving model to weights\weights_DNN_1.hdf5
Epoch 21/100
- 79s - loss: 0.1143 - acc: 0.9725 - val_loss: 2.1726 - val_acc: 0.7339
Epoch 00021: val_acc did not improve from 0.74185
Epoch 22/100
- 33s - loss: 0.1017 - acc: 0.9775 - val_loss: 2.2795 - val_acc: 0.7373
Epoch 00022: val_acc did not improve from 0.74185
Epoch 23/100
- 41s - loss: 0.1006 - acc: 0.9787 - val_loss: 2.3428 - val_acc: 0.7289
Epoch 00023: val_acc did not improve from 0.74185
Epoch 24/100
- 98s - loss: 0.0889 - acc: 0.9797 - val_loss: 2.3089 - val_acc: 0.7314
Epoch 00024: val_acc did not improve from 0.74185
Epoch 25/100
- 42s - loss: 0.0940 - acc: 0.9794 - val_loss: 2.2587 - val_acc: 0.7343
Epoch 00025: val_acc did not improve from 0.74185
Epoch 26/100
- 34s - loss: 0.0884 - acc: 0.9808 - val_loss: 2.3147 - val_acc: 0.7352
Epoch 00026: val_acc did not improve from 0.74185
Epoch 27/100
- 32s - loss: 0.0805 - acc: 0.9827 - val_loss: 2.4054 - val_acc: 0.7435
Epoch 00027: val_acc improved from 0.74185 to 0.74353, saving model to weights\weights_DNN_1.hdf5
Epoch 28/100
- 76s - loss: 0.0854 - acc: 0.9829 - val_loss: 2.4696 - val_acc: 0.7356
Epoch 00028: val_acc did not improve from 0.74353
Epoch 29/100
- 76s - loss: 0.0898 - acc: 0.9832 - val_loss: 2.4101 - val_acc: 0.7364
Epoch 00029: val_acc did not improve from 0.74353
Epoch 30/100
- 70s - loss: 0.0845 - acc: 0.9827 - val_loss: 2.4140 - val_acc: 0.7352
Epoch 00030: val_acc did not improve from 0.74353
Epoch 31/100
- 83s - loss: 0.0887 - acc: 0.9837 - val_loss: 2.3476 - val_acc: 0.7444
Epoch 00031: val_acc improved from 0.74353 to 0.74436, saving model to weights\weights_DNN_1.hdf5
Epoch 32/100
- 45s - loss: 0.0744 - acc: 0.9844 - val_loss: 2.4697 - val_acc: 0.7381
...
Epoch 00098: val_acc did not improve from 0.74436
Epoch 99/100
- 76s - loss: 0.0722 - acc: 0.9921 - val_loss: 2.7782 - val_acc: 0.7285
Epoch 00099: val_acc did not improve from 0.74436
Epoch 100/100
- 82s - loss: 0.0726 - acc: 0.9929 - val_loss: 2.7103 - val_acc: 0.7306
Epoch 00100: val_acc did not improve from 0.74436
DNN 2
<keras.optimizers.Adagrad object at 0x7fefbe1ffda0>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 126s - loss: 3.3523 - acc: 0.0518 - val_loss: 3.0346 - val_acc: 0.0794
Epoch 00001: val_acc improved from -inf to 0.07937, saving model to weights\weights_DNN_2.hdf5
Epoch 2/100
- 92s - loss: 2.8354 - acc: 0.1086 - val_loss: 2.5724 - val_acc: 0.1612
Epoch 00002: val_acc improved from 0.07937 to 0.16124, saving model to weights\weights_DNN_2.hdf5
Epoch 3/100
- 83s - loss: 2.4536 - acc: 0.1649 - val_loss: 2.3181 - val_acc: 0.2264
Epoch 00003: val_acc improved from 0.16124 to 0.22640, saving model to weights\weights_DNN_2.hdf5
Epoch 4/100
- 113s - loss: 2.1812 - acc: 0.2184 - val_loss: 2.1923 - val_acc: 0.2715
Epoch 00004: val_acc improved from 0.22640 to 0.27151, saving model to weights\weights_DNN_2.hdf5
Epoch 5/100
- 32s - loss: 1.9945 - acc: 0.2637 - val_loss: 2.0933 - val_acc: 0.2790
Epoch 00005: val_acc improved from 0.27151 to 0.27903, saving model to weights\weights_DNN_2.hdf5
Epoch 6/100
- 117s - loss: 1.8165 - acc: 0.3100 - val_loss: 1.9676 - val_acc: 0.3425
Epoch 00006: val_acc improved from 0.27903 to 0.34252, saving model to weights\weights_DNN_2.hdf5
Epoch 7/100
- 90s - loss: 1.6621 - acc: 0.3679 - val_loss: 1.8871 - val_acc: 0.3743
Epoch 00007: val_acc improved from 0.34252 to 0.37427, saving model to weights\weights_DNN_2.hdf5
Epoch 8/100
- 155s - loss: 1.5050 - acc: 0.4136 - val_loss: 1.8406 - val_acc: 0.4215
Epoch 00008: val_acc improved from 0.37427 to 0.42147, saving model to weights\weights_DNN_2.hdf5
Epoch 9/100
- 188s - loss: 1.3478 - acc: 0.4779 - val_loss: 1.7912 - val_acc: 0.4390
Epoch 00009: val_acc improved from 0.42147 to 0.43901, saving model to weights\weights_DNN_2.hdf5
Epoch 10/100
- 131s - loss: 1.2100 - acc: 0.5168 - val_loss: 1.7741 - val_acc: 0.4616
Epoch 00010: val_acc improved from 0.43901 to 0.46157, saving model to weights\weights_DNN_2.hdf5
Epoch 11/100
- 160s - loss: 1.1056 - acc: 0.5582 - val_loss: 1.8189 - val_acc: 0.4595
Epoch 00011: val_acc did not improve from 0.46157
Epoch 12/100
- 147s - loss: 1.0053 - acc: 0.5919 - val_loss: 1.8520 - val_acc: 0.4799
Epoch 00012: val_acc improved from 0.46157 to 0.47995, saving model to weights\weights_DNN_2.hdf5
Epoch 13/100
- 126s - loss: 0.9275 - acc: 0.6248 - val_loss: 1.8868 - val_acc: 0.5063
Epoch 00013: val_acc improved from 0.47995 to 0.50627, saving model to weights\weights_DNN_2.hdf5
Epoch 14/100
- 164s - loss: 0.8626 - acc: 0.6500 - val_loss: 1.8809 - val_acc: 0.5209
Epoch 00014: val_acc improved from 0.50627 to 0.52089, saving model to weights\weights_DNN_2.hdf5
Epoch 15/100
- 90s - loss: 0.8143 - acc: 0.6720 - val_loss: 1.9335 - val_acc: 0.5439
Epoch 00015: val_acc improved from 0.52089 to 0.54386, saving model to weights\weights_DNN_2.hdf5
Epoch 16/100
- 105s - loss: 0.7603 - acc: 0.6953 - val_loss: 1.9360 - val_acc: 0.5422
Epoch 00016: val_acc did not improve from 0.54386
Epoch 17/100
- 139s - loss: 0.7173 - acc: 0.7124 - val_loss: 1.9910 - val_acc: 0.5514
Epoch 00017: val_acc improved from 0.54386 to 0.55138, saving model to weights\weights_DNN_2.hdf5
Epoch 18/100
- 221s - loss: 0.6764 - acc: 0.7383 - val_loss: 1.9660 - val_acc: 0.5560
Epoch 00018: val_acc improved from 0.55138 to 0.55597, saving model to weights\weights_DNN_2.hdf5
Epoch 19/100
- 80s - loss: 0.6203 - acc: 0.7607 - val_loss: 2.0173 - val_acc: 0.5668
Epoch 00019: val_acc improved from 0.55597 to 0.56683, saving model to weights\weights_DNN_2.hdf5
Epoch 20/100
- 159s - loss: 0.5948 - acc: 0.7719 - val_loss: 2.0632 - val_acc: 0.5631
Epoch 00020: val_acc did not improve from 0.56683
Epoch 21/100
- 91s - loss: 0.5639 - acc: 0.7848 - val_loss: 2.0946 - val_acc: 0.5714
Epoch 00021: val_acc improved from 0.56683 to 0.57143, saving model to weights\weights_DNN_2.hdf5
Epoch 22/100
- 69s - loss: 0.5291 - acc: 0.8010 - val_loss: 2.1178 - val_acc: 0.5752
Epoch 00022: val_acc improved from 0.57143 to 0.57519, saving model to weights\weights_DNN_2.hdf5
Epoch 23/100
- 35s - loss: 0.5034 - acc: 0.8144 - val_loss: 2.1444 - val_acc: 0.5714
Epoch 00023: val_acc did not improve from 0.57519
Epoch 24/100
- 94s - loss: 0.4700 - acc: 0.8277 - val_loss: 2.1417 - val_acc: 0.5840
Epoch 00024: val_acc improved from 0.57519 to 0.58396, saving model to weights\weights_DNN_2.hdf5
Epoch 25/100
- 114s - loss: 0.4330 - acc: 0.8405 - val_loss: 2.1789 - val_acc: 0.5965
Epoch 00025: val_acc improved from 0.58396 to 0.59649, saving model to weights\weights_DNN_2.hdf5
Epoch 26/100
- 119s - loss: 0.4204 - acc: 0.8481 - val_loss: 2.2221 - val_acc: 0.6036
Epoch 00026: val_acc improved from 0.59649 to 0.60359, saving model to weights\weights_DNN_2.hdf5
Epoch 27/100
- 69s - loss: 0.4015 - acc: 0.8578 - val_loss: 2.2306 - val_acc: 0.6048
Epoch 00027: val_acc improved from 0.60359 to 0.60485, saving model to weights\weights_DNN_2.hdf5
Epoch 28/100
- 86s - loss: 0.3688 - acc: 0.8693 - val_loss: 2.2779 - val_acc: 0.6036
Epoch 00028: val_acc did not improve from 0.60485
Epoch 29/100
- 97s - loss: 0.3589 - acc: 0.8758 - val_loss: 2.3185 - val_acc: 0.6003
Epoch 00029: val_acc did not improve from 0.60485
Epoch 30/100
- 121s - loss: 0.3477 - acc: 0.8762 - val_loss: 2.3634 - val_acc: 0.6082
Epoch 00030: val_acc improved from 0.60485 to 0.60819, saving model to weights\weights_DNN_2.hdf5
Epoch 31/100
- 122s - loss: 0.3255 - acc: 0.8801 - val_loss: 2.3722 - val_acc: 0.6132
Epoch 00031: val_acc improved from 0.60819 to 0.61320, saving model to weights\weights_DNN_2.hdf5
Epoch 32/100
- 142s - loss: 0.3101 - acc: 0.8918 - val_loss: 2.4105 - val_acc: 0.6053
Epoch 00032: val_acc did not improve from 0.61320
Epoch 33/100
- 141s - loss: 0.2980 - acc: 0.8980 - val_loss: 2.3923 - val_acc: 0.6241
Epoch 00033: val_acc improved from 0.61320 to 0.62406, saving model to weights\weights_DNN_2.hdf5
Epoch 34/100
- 177s - loss: 0.2880 - acc: 0.9006 - val_loss: 2.3945 - val_acc: 0.6266
Epoch 00034: val_acc improved from 0.62406 to 0.62657, saving model to weights\weights_DNN_2.hdf5
Epoch 35/100
- 148s - loss: 0.2819 - acc: 0.9014 - val_loss: 2.4723 - val_acc: 0.6157
Epoch 00035: val_acc did not improve from 0.62657
Epoch 36/100
- 182s - loss: 0.2708 - acc: 0.9076 - val_loss: 2.4338 - val_acc: 0.6249
Epoch 00036: val_acc did not improve from 0.62657
Epoch 37/100
- 112s - loss: 0.2539 - acc: 0.9135 - val_loss: 2.4850 - val_acc: 0.6232
Epoch 00037: val_acc did not improve from 0.62657
Epoch 38/100
- 109s - loss: 0.2609 - acc: 0.9120 - val_loss: 2.4761 - val_acc: 0.6312
Epoch 00038: val_acc improved from 0.62657 to 0.63116, saving model to weights\weights_DNN_2.hdf5
Epoch 39/100
- 48s - loss: 0.2373 - acc: 0.9185 - val_loss: 2.5396 - val_acc: 0.6241
Epoch 00039: val_acc did not improve from 0.63116
Epoch 40/100
- 44s - loss: 0.2302 - acc: 0.9224 - val_loss: 2.5320 - val_acc: 0.6295
Epoch 00040: val_acc did not improve from 0.63116
Epoch 41/100
- 87s - loss: 0.2303 - acc: 0.9227 - val_loss: 2.5506 - val_acc: 0.6278
Epoch 00041: val_acc did not improve from 0.63116
Epoch 42/100
- 15s - loss: 0.2240 - acc: 0.9254 - val_loss: 2.5231 - val_acc: 0.6358
Epoch 00042: val_acc improved from 0.63116 to 0.63576, saving model to weights\weights_DNN_2.hdf5
Epoch 43/100
- 31s - loss: 0.2166 - acc: 0.9269 - val_loss: 2.5401 - val_acc: 0.6353
Epoch 00043: val_acc did not improve from 0.63576
Epoch 44/100
- 17s - loss: 0.2068 - acc: 0.9294 - val_loss: 2.5487 - val_acc: 0.6412
Epoch 00044: val_acc improved from 0.63576 to 0.64119, saving model to weights\weights_DNN_2.hdf5
Epoch 45/100
- 35s - loss: 0.1918 - acc: 0.9360 - val_loss: 2.5554 - val_acc: 0.6504
Epoch 00045: val_acc improved from 0.64119 to 0.65038, saving model to weights\weights_DNN_2.hdf5
Epoch 46/100
- 25s - loss: 0.1871 - acc: 0.9382 - val_loss: 2.5567 - val_acc: 0.6512
Epoch 00046: val_acc improved from 0.65038 to 0.65121, saving model to weights\weights_DNN_2.hdf5
Epoch 47/100
- 21s - loss: 0.1833 - acc: 0.9409 - val_loss: 2.6156 - val_acc: 0.6508
Epoch 00047: val_acc did not improve from 0.65121
Epoch 48/100
- 31s - loss: 0.1898 - acc: 0.9389 - val_loss: 2.6360 - val_acc: 0.6458
Epoch 00048: val_acc did not improve from 0.65121
Epoch 49/100
- 22s - loss: 0.1756 - acc: 0.9436 - val_loss: 2.5959 - val_acc: 0.6566
Epoch 00049: val_acc improved from 0.65121 to 0.65664, saving model to weights\weights_DNN_2.hdf5
Epoch 50/100
- 39s - loss: 0.1647 - acc: 0.9494 - val_loss: 2.6494 - val_acc: 0.6500
Epoch 00050: val_acc did not improve from 0.65664
Epoch 51/100
- 37s - loss: 0.1604 - acc: 0.9503 - val_loss: 2.6780 - val_acc: 0.6550
Epoch 00051: val_acc did not improve from 0.65664
Epoch 52/100
- 26s - loss: 0.1681 - acc: 0.9487 - val_loss: 2.6621 - val_acc: 0.6537
Epoch 00052: val_acc did not improve from 0.65664
Epoch 53/100
- 24s - loss: 0.1642 - acc: 0.9494 - val_loss: 2.6853 - val_acc: 0.6520
Epoch 00053: val_acc did not improve from 0.65664
Epoch 54/100
- 29s - loss: 0.1510 - acc: 0.9489 - val_loss: 2.7013 - val_acc: 0.6512
Epoch 00054: val_acc did not improve from 0.65664
Epoch 55/100
- 24s - loss: 0.1411 - acc: 0.9563 - val_loss: 2.6764 - val_acc: 0.6662
Epoch 00055: val_acc improved from 0.65664 to 0.66625, saving model to weights\weights_DNN_2.hdf5
Epoch 56/100
- 22s - loss: 0.1367 - acc: 0.9574 - val_loss: 2.6647 - val_acc: 0.6604
Epoch 00056: val_acc did not improve from 0.66625
Epoch 57/100
- 19s - loss: 0.1376 - acc: 0.9561 - val_loss: 2.6930 - val_acc: 0.6617
Epoch 00057: val_acc did not improve from 0.66625
Epoch 58/100
- 18s - loss: 0.1249 - acc: 0.9600 - val_loss: 2.7008 - val_acc: 0.6629
Epoch 00058: val_acc did not improve from 0.66625
Epoch 59/100
- 20s - loss: 0.1265 - acc: 0.9615 - val_loss: 2.7126 - val_acc: 0.6667
Epoch 00059: val_acc improved from 0.66625 to 0.66667, saving model to weights\weights_DNN_2.hdf5
Epoch 60/100
- 45s - loss: 0.1180 - acc: 0.9636 - val_loss: 2.6867 - val_acc: 0.6717
Epoch 00060: val_acc improved from 0.66667 to 0.67168, saving model to weights\weights_DNN_2.hdf5
Epoch 61/100
- 39s - loss: 0.1135 - acc: 0.9656 - val_loss: 2.7265 - val_acc: 0.6646
Epoch 00061: val_acc did not improve from 0.67168
Epoch 62/100
- 33s - loss: 0.1266 - acc: 0.9624 - val_loss: 2.7022 - val_acc: 0.6700
Epoch 00062: val_acc did not improve from 0.67168
Epoch 63/100
- 26s - loss: 0.1124 - acc: 0.9676 - val_loss: 2.7556 - val_acc: 0.6729
Epoch 00063: val_acc improved from 0.67168 to 0.67293, saving model to weights\weights_DNN_2.hdf5
Epoch 64/100
- 34s - loss: 0.1189 - acc: 0.9650 - val_loss: 2.7080 - val_acc: 0.6750
Epoch 00064: val_acc improved from 0.67293 to 0.67502, saving model to weights\weights_DNN_2.hdf5
Epoch 65/100
- 39s - loss: 0.1054 - acc: 0.9676 - val_loss: 2.7325 - val_acc: 0.6784
Epoch 00065: val_acc improved from 0.67502 to 0.67836, saving model to weights\weights_DNN_2.hdf5
Epoch 66/100
- 34s - loss: 0.1042 - acc: 0.9695 - val_loss: 2.7302 - val_acc: 0.6746
Epoch 00066: val_acc did not improve from 0.67836
Epoch 67/100
- 35s - loss: 0.1024 - acc: 0.9674 - val_loss: 2.7585 - val_acc: 0.6742
Epoch 00067: val_acc did not improve from 0.67836
Epoch 68/100
- 37s - loss: 0.1049 - acc: 0.9696 - val_loss: 2.7348 - val_acc: 0.6734
Epoch 00068: val_acc did not improve from 0.67836
Epoch 69/100
- 39s - loss: 0.1112 - acc: 0.9665 - val_loss: 2.7468 - val_acc: 0.6713
Epoch 00069: val_acc did not improve from 0.67836
Epoch 70/100
- 37s - loss: 0.0905 - acc: 0.9715 - val_loss: 2.7337 - val_acc: 0.6809
Epoch 00070: val_acc improved from 0.67836 to 0.68087, saving model to weights\weights_DNN_2.hdf5
Epoch 71/100
- 32s - loss: 0.0912 - acc: 0.9710 - val_loss: 2.7616 - val_acc: 0.6800
Epoch 00071: val_acc did not improve from 0.68087
Epoch 72/100
- 36s - loss: 0.0944 - acc: 0.9703 - val_loss: 2.7593 - val_acc: 0.6825
Epoch 00072: val_acc improved from 0.68087 to 0.68254, saving model to weights\weights_DNN_2.hdf5
Epoch 73/100
- 31s - loss: 0.0889 - acc: 0.9735 - val_loss: 2.7266 - val_acc: 0.6846
Epoch 00073: val_acc improved from 0.68254 to 0.68463, saving model to weights\weights_DNN_2.hdf5
Epoch 74/100
- 32s - loss: 0.0841 - acc: 0.9767 - val_loss: 2.8209 - val_acc: 0.6759
Epoch 00074: val_acc did not improve from 0.68463
Epoch 75/100
- 36s - loss: 0.0778 - acc: 0.9759 - val_loss: 2.7884 - val_acc: 0.6867
Epoch 00075: val_acc improved from 0.68463 to 0.68672, saving model to weights\weights_DNN_2.hdf5
Epoch 76/100
- 50s - loss: 0.0806 - acc: 0.9757 - val_loss: 2.8117 - val_acc: 0.6813
Epoch 00076: val_acc did not improve from 0.68672
Epoch 77/100
- 52s - loss: 0.0820 - acc: 0.9752 - val_loss: 2.7971 - val_acc: 0.6859
Epoch 00077: val_acc did not improve from 0.68672
Epoch 78/100
- 28s - loss: 0.0708 - acc: 0.9798 - val_loss: 2.8323 - val_acc: 0.6834
Epoch 00078: val_acc did not improve from 0.68672
Epoch 79/100
- 38s - loss: 0.0726 - acc: 0.9793 - val_loss: 2.8103 - val_acc: 0.6876
Epoch 00079: val_acc improved from 0.68672 to 0.68755, saving model to weights\weights_DNN_2.hdf5
Epoch 80/100
- 30s - loss: 0.0825 - acc: 0.9768 - val_loss: 2.7935 - val_acc: 0.6905
Epoch 00080: val_acc improved from 0.68755 to 0.69048, saving model to weights\weights_DNN_2.hdf5
Epoch 81/100
- 43s - loss: 0.0753 - acc: 0.9788 - val_loss: 2.8216 - val_acc: 0.6850
Epoch 00081: val_acc did not improve from 0.69048
Epoch 82/100
- 30s - loss: 0.0800 - acc: 0.9772 - val_loss: 2.7802 - val_acc: 0.6921
Epoch 00082: val_acc improved from 0.69048 to 0.69215, saving model to weights\weights_DNN_2.hdf5
Epoch 83/100
- 42s - loss: 0.0706 - acc: 0.9811 - val_loss: 2.8115 - val_acc: 0.6867
Epoch 00083: val_acc did not improve from 0.69215
Epoch 84/100
- 23s - loss: 0.0680 - acc: 0.9823 - val_loss: 2.8257 - val_acc: 0.6876
Epoch 00084: val_acc did not improve from 0.69215
Epoch 85/100
- 17s - loss: 0.0732 - acc: 0.9796 - val_loss: 2.7800 - val_acc: 0.6921
Epoch 00085: val_acc did not improve from 0.69215
Epoch 86/100
- 14s - loss: 0.0675 - acc: 0.9795 - val_loss: 2.7615 - val_acc: 0.6967
Epoch 00086: val_acc improved from 0.69215 to 0.69674, saving model to weights\weights_DNN_2.hdf5
Epoch 87/100
- 15s - loss: 0.0583 - acc: 0.9836 - val_loss: 2.7877 - val_acc: 0.6959
Epoch 00087: val_acc did not improve from 0.69674
Epoch 88/100
- 20s - loss: 0.0637 - acc: 0.9813 - val_loss: 2.8376 - val_acc: 0.6934
Epoch 00088: val_acc did not improve from 0.69674
Epoch 89/100
- 12s - loss: 0.0547 - acc: 0.9835 - val_loss: 2.8297 - val_acc: 0.6942
Epoch 00089: val_acc did not improve from 0.69674
Epoch 90/100
- 13s - loss: 0.0551 - acc: 0.9851 - val_loss: 2.8544 - val_acc: 0.6951
Epoch 00090: val_acc did not improve from 0.69674
Epoch 91/100
- 8s - loss: 0.0627 - acc: 0.9838 - val_loss: 2.8320 - val_acc: 0.6976
Epoch 00091: val_acc improved from 0.69674 to 0.69758, saving model to weights\weights_DNN_2.hdf5
Epoch 92/100
- 8s - loss: 0.0474 - acc: 0.9863 - val_loss: 2.8458 - val_acc: 0.6980
Epoch 00092: val_acc improved from 0.69758 to 0.69799, saving model to weights\weights_DNN_2.hdf5
Epoch 93/100
- 17s - loss: 0.0536 - acc: 0.9856 - val_loss: 2.8812 - val_acc: 0.6947
Epoch 00093: val_acc did not improve from 0.69799
Epoch 94/100
- 8s - loss: 0.0647 - acc: 0.9844 - val_loss: 2.8945 - val_acc: 0.6909
Epoch 00094: val_acc did not improve from 0.69799
Epoch 95/100
- 9s - loss: 0.0518 - acc: 0.9836 - val_loss: 2.8468 - val_acc: 0.6972
Epoch 00095: val_acc did not improve from 0.69799
Epoch 96/100
- 10s - loss: 0.0608 - acc: 0.9843 - val_loss: 2.8292 - val_acc: 0.6976
Epoch 00096: val_acc did not improve from 0.69799
Epoch 97/100
- 11s - loss: 0.0481 - acc: 0.9872 - val_loss: 2.8654 - val_acc: 0.6967
Epoch 00097: val_acc did not improve from 0.69799
Epoch 98/100
- 15s - loss: 0.0488 - acc: 0.9862 - val_loss: 2.8260 - val_acc: 0.6997
Epoch 00098: val_acc improved from 0.69799 to 0.69967, saving model to weights\weights_DNN_2.hdf5
Epoch 99/100
- 14s - loss: 0.0499 - acc: 0.9864 - val_loss: 2.9037 - val_acc: 0.6959
Epoch 00099: val_acc did not improve from 0.69967
Epoch 100/100
- 14s - loss: 0.0525 - acc: 0.9865 - val_loss: 2.8700 - val_acc: 0.7059
Epoch 00100: val_acc improved from 0.69967 to 0.70593, saving model to weights\weights_DNN_2.hdf5
RNN 0
3
42
<keras.optimizers.RMSprop object at 0x7ff0e4d8aa20>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 691s - loss: 3.2295 - acc: 0.0733 - val_loss: 2.8583 - val_acc: 0.1266
Epoch 00001: val_acc improved from -inf to 0.12657, saving model to weights\weights_RNN_0.hdf5
Epoch 2/100
- 707s - loss: 2.7057 - acc: 0.1271 - val_loss: 2.5021 - val_acc: 0.1725
Epoch 00002: val_acc improved from 0.12657 to 0.17251, saving model to weights\weights_RNN_0.hdf5
Epoch 3/100
- 619s - loss: 2.4545 - acc: 0.1579 - val_loss: 2.3607 - val_acc: 0.1621
Epoch 00003: val_acc did not improve from 0.17251
Epoch 4/100
- 584s - loss: 2.2792 - acc: 0.1967 - val_loss: 2.2207 - val_acc: 0.2118
Epoch 00004: val_acc improved from 0.17251 to 0.21178, saving model to weights\weights_RNN_0.hdf5
Epoch 5/100
- 604s - loss: 2.1011 - acc: 0.2431 - val_loss: 2.1073 - val_acc: 0.2707
Epoch 00005: val_acc improved from 0.21178 to 0.27068, saving model to weights\weights_RNN_0.hdf5
Epoch 6/100
- 584s - loss: 1.9101 - acc: 0.2998 - val_loss: 1.9260 - val_acc: 0.3329
Epoch 00006: val_acc improved from 0.27068 to 0.33292, saving model to weights\weights_RNN_0.hdf5
Epoch 7/100
- 599s - loss: 1.7370 - acc: 0.3543 - val_loss: 1.8417 - val_acc: 0.3642
Epoch 00007: val_acc improved from 0.33292 to 0.36424, saving model to weights\weights_RNN_0.hdf5
Epoch 8/100
- 581s - loss: 1.5715 - acc: 0.4252 - val_loss: 1.6597 - val_acc: 0.4407
Epoch 00008: val_acc improved from 0.36424 to 0.44069, saving model to weights\weights_RNN_0.hdf5
Epoch 9/100
- 593s - loss: 1.4288 - acc: 0.4817 - val_loss: 1.5140 - val_acc: 0.5138
Epoch 00009: val_acc improved from 0.44069 to 0.51378, saving model to weights\weights_RNN_0.hdf5
Epoch 10/100
- 600s - loss: 1.2731 - acc: 0.5410 - val_loss: 1.4982 - val_acc: 0.5263
Epoch 00010: val_acc improved from 0.51378 to 0.52632, saving model to weights\weights_RNN_0.hdf5
Epoch 11/100
- 599s - loss: 1.1504 - acc: 0.5878 - val_loss: 1.4324 - val_acc: 0.5660
Epoch 00011: val_acc improved from 0.52632 to 0.56600, saving model to weights\weights_RNN_0.hdf5
Epoch 12/100
- 591s - loss: 1.0231 - acc: 0.6374 - val_loss: 1.5650 - val_acc: 0.5526
Epoch 00012: val_acc did not improve from 0.56600
Epoch 13/100
- 605s - loss: 0.9246 - acc: 0.6751 - val_loss: 1.4499 - val_acc: 0.6078
Epoch 00013: val_acc improved from 0.56600 to 0.60777, saving model to weights\weights_RNN_0.hdf5
Epoch 14/100
- 590s - loss: 0.8146 - acc: 0.7195 - val_loss: 1.4465 - val_acc: 0.6245
Epoch 00014: val_acc improved from 0.60777 to 0.62448, saving model to weights\weights_RNN_0.hdf5
Epoch 15/100
- 598s - loss: 0.7233 - acc: 0.7537 - val_loss: 1.4578 - val_acc: 0.6241
Epoch 00015: val_acc did not improve from 0.62448
Epoch 16/100
- 583s - loss: 0.6361 - acc: 0.7875 - val_loss: 1.4714 - val_acc: 0.6587
Epoch 00016: val_acc improved from 0.62448 to 0.65873, saving model to weights\weights_RNN_0.hdf5
Epoch 17/100
- 610s - loss: 0.5822 - acc: 0.8051 - val_loss: 1.4878 - val_acc: 0.6667
Epoch 00017: val_acc improved from 0.65873 to 0.66667, saving model to weights\weights_RNN_0.hdf5
Epoch 18/100
- 580s - loss: 0.5164 - acc: 0.8273 - val_loss: 1.5833 - val_acc: 0.6399
Epoch 00018: val_acc did not improve from 0.66667
Epoch 19/100
- 576s - loss: 0.4609 - acc: 0.8482 - val_loss: 1.5653 - val_acc: 0.6713
Epoch 00019: val_acc improved from 0.66667 to 0.67126, saving model to weights\weights_RNN_0.hdf5
Epoch 20/100
- 611s - loss: 0.4144 - acc: 0.8636 - val_loss: 1.6817 - val_acc: 0.6742
Epoch 00020: val_acc improved from 0.67126 to 0.67419, saving model to weights\weights_RNN_0.hdf5
Epoch 21/100
- 570s - loss: 0.3633 - acc: 0.8840 - val_loss: 1.8648 - val_acc: 0.6454
Epoch 00021: val_acc did not improve from 0.67419
Epoch 22/100
- 583s - loss: 0.3363 - acc: 0.8917 - val_loss: 1.7600 - val_acc: 0.6742
Epoch 00022: val_acc did not improve from 0.67419
Epoch 23/100
- 603s - loss: 0.3020 - acc: 0.9056 - val_loss: 1.8093 - val_acc: 0.6746
Epoch 00023: val_acc improved from 0.67419 to 0.67460, saving model to weights\weights_RNN_0.hdf5
Epoch 24/100
- 577s - loss: 0.2714 - acc: 0.9106 - val_loss: 1.8019 - val_acc: 0.6888
Epoch 00024: val_acc improved from 0.67460 to 0.68881, saving model to weights\weights_RNN_0.hdf5
Epoch 25/100
- 585s - loss: 0.2474 - acc: 0.9190 - val_loss: 1.9557 - val_acc: 0.6763
Epoch 00025: val_acc did not improve from 0.68881
Epoch 26/100
- 562s - loss: 0.2184 - acc: 0.9326 - val_loss: 2.0493 - val_acc: 0.6646
Epoch 00026: val_acc did not improve from 0.68881
Epoch 27/100
- 569s - loss: 0.2041 - acc: 0.9352 - val_loss: 2.0344 - val_acc: 0.6834
Epoch 00027: val_acc did not improve from 0.68881
Epoch 28/100
- 589s - loss: 0.1897 - acc: 0.9397 - val_loss: 1.9704 - val_acc: 0.6967
Epoch 00028: val_acc improved from 0.68881 to 0.69674, saving model to weights\weights_RNN_0.hdf5
Epoch 29/100
- 572s - loss: 0.1689 - acc: 0.9452 - val_loss: 2.2266 - val_acc: 0.6800
Epoch 00029: val_acc did not improve from 0.69674
Epoch 30/100
- 550s - loss: 0.1560 - acc: 0.9530 - val_loss: 2.1155 - val_acc: 0.6984
Epoch 00030: val_acc improved from 0.69674 to 0.69841, saving model to weights\weights_RNN_0.hdf5
Epoch 31/100
- 543s - loss: 0.1448 - acc: 0.9552 - val_loss: 2.3955 - val_acc: 0.6546
Epoch 00031: val_acc did not improve from 0.69841
Epoch 32/100
- 577s - loss: 0.1385 - acc: 0.9570 - val_loss: 2.2567 - val_acc: 0.6871
Epoch 00032: val_acc did not improve from 0.69841
Epoch 33/100
- 578s - loss: 0.1224 - acc: 0.9610 - val_loss: 2.5516 - val_acc: 0.6679
Epoch 00033: val_acc did not improve from 0.69841
Epoch 34/100
- 615s - loss: 0.1280 - acc: 0.9619 - val_loss: 2.2845 - val_acc: 0.6963
Epoch 00034: val_acc did not improve from 0.69841
Epoch 35/100
- 689s - loss: 0.1208 - acc: 0.9623 - val_loss: 2.4526 - val_acc: 0.6805
Epoch 00035: val_acc did not improve from 0.69841
Epoch 36/100
- 763s - loss: 0.1143 - acc: 0.9664 - val_loss: 2.4973 - val_acc: 0.6742
Epoch 00036: val_acc did not improve from 0.69841
Epoch 37/100
- 762s - loss: 0.1065 - acc: 0.9672 - val_loss: 2.4875 - val_acc: 0.6909
Epoch 00037: val_acc did not improve from 0.69841
Epoch 38/100
- 776s - loss: 0.0977 - acc: 0.9714 - val_loss: 2.5168 - val_acc: 0.6884
Epoch 00038: val_acc did not improve from 0.69841
Epoch 39/100
- 710s - loss: 0.0927 - acc: 0.9705 - val_loss: 2.8769 - val_acc: 0.6662
Epoch 00039: val_acc did not improve from 0.69841
Epoch 40/100
- 713s - loss: 0.0831 - acc: 0.9739 - val_loss: 2.5925 - val_acc: 0.6909
Epoch 00040: val_acc did not improve from 0.69841
Epoch 41/100
- 700s - loss: 0.0797 - acc: 0.9746 - val_loss: 2.7547 - val_acc: 0.6784
Epoch 00041: val_acc did not improve from 0.69841
Epoch 42/100
- 678s - loss: 0.0752 - acc: 0.9756 - val_loss: 2.7027 - val_acc: 0.6909
Epoch 00042: val_acc did not improve from 0.69841
Epoch 43/100
- 715s - loss: 0.0793 - acc: 0.9776 - val_loss: 2.8076 - val_acc: 0.6867
Epoch 00043: val_acc did not improve from 0.69841
Epoch 44/100
- 678s - loss: 0.0664 - acc: 0.9795 - val_loss: 2.8500 - val_acc: 0.6842
Epoch 00044: val_acc did not improve from 0.69841
Epoch 45/100
- 691s - loss: 0.0720 - acc: 0.9783 - val_loss: 3.2216 - val_acc: 0.6525
Epoch 00045: val_acc did not improve from 0.69841
Epoch 46/100
- 673s - loss: 0.0678 - acc: 0.9802 - val_loss: 2.9903 - val_acc: 0.6821
Epoch 00046: val_acc did not improve from 0.69841
Epoch 47/100
- 661s - loss: 0.0617 - acc: 0.9806 - val_loss: 2.9695 - val_acc: 0.6913
Epoch 00047: val_acc did not improve from 0.69841
Epoch 48/100
- 691s - loss: 0.0604 - acc: 0.9820 - val_loss: 3.1651 - val_acc: 0.6759
Epoch 00048: val_acc did not improve from 0.69841
Epoch 49/100
- 709s - loss: 0.0634 - acc: 0.9799 - val_loss: 3.1216 - val_acc: 0.6813
Epoch 00049: val_acc did not improve from 0.69841
Epoch 50/100
- 696s - loss: 0.0562 - acc: 0.9841 - val_loss: 2.9634 - val_acc: 0.7009
Epoch 00050: val_acc improved from 0.69841 to 0.70092, saving model to weights\weights_RNN_0.hdf5
Epoch 51/100
- 705s - loss: 0.0597 - acc: 0.9838 - val_loss: 3.1089 - val_acc: 0.6867
Epoch 00051: val_acc did not improve from 0.70092
Epoch 52/100
- 641s - loss: 0.0550 - acc: 0.9823 - val_loss: 3.3981 - val_acc: 0.6692
Epoch 00052: val_acc did not improve from 0.70092
Epoch 53/100
- 566s - loss: 0.0459 - acc: 0.9847 - val_loss: 3.1833 - val_acc: 0.6813
Epoch 00053: val_acc did not improve from 0.70092
Epoch 54/100
- 550s - loss: 0.0491 - acc: 0.9850 - val_loss: 3.0295 - val_acc: 0.7055
Epoch 00054: val_acc improved from 0.70092 to 0.70551, saving model to weights\weights_RNN_0.hdf5
Epoch 55/100
- 500s - loss: 0.0504 - acc: 0.9856 - val_loss: 3.1576 - val_acc: 0.6955
Epoch 00055: val_acc did not improve from 0.70551
Epoch 56/100
- 489s - loss: 0.0522 - acc: 0.9835 - val_loss: 2.9685 - val_acc: 0.7068
Epoch 00056: val_acc improved from 0.70551 to 0.70677, saving model to weights\weights_RNN_0.hdf5
Epoch 57/100
- 482s - loss: 0.0510 - acc: 0.9851 - val_loss: 3.0108 - val_acc: 0.7051
Epoch 00057: val_acc did not improve from 0.70677
Epoch 58/100
- 524s - loss: 0.0523 - acc: 0.9850 - val_loss: 3.1915 - val_acc: 0.6926
Epoch 00058: val_acc did not improve from 0.70677
Epoch 59/100
- 546s - loss: 0.0494 - acc: 0.9843 - val_loss: 3.2815 - val_acc: 0.6763
Epoch 00059: val_acc did not improve from 0.70677
Epoch 60/100
- 545s - loss: 0.0454 - acc: 0.9852 - val_loss: 3.3623 - val_acc: 0.6759
Epoch 00060: val_acc did not improve from 0.70677
Epoch 61/100
- 547s - loss: 0.0422 - acc: 0.9863 - val_loss: 3.4174 - val_acc: 0.6679
Epoch 00061: val_acc did not improve from 0.70677
Epoch 62/100
- 537s - loss: 0.0448 - acc: 0.9867 - val_loss: 3.3015 - val_acc: 0.6892
Epoch 00062: val_acc did not improve from 0.70677
Epoch 63/100
- 548s - loss: 0.0462 - acc: 0.9854 - val_loss: 3.5555 - val_acc: 0.6654
Epoch 00063: val_acc did not improve from 0.70677
Epoch 64/100
- 545s - loss: 0.0458 - acc: 0.9870 - val_loss: 3.3253 - val_acc: 0.6871
Epoch 00064: val_acc did not improve from 0.70677
Epoch 65/100
- 532s - loss: 0.0474 - acc: 0.9872 - val_loss: 3.3957 - val_acc: 0.6779
Epoch 00065: val_acc did not improve from 0.70677
Epoch 66/100
- 548s - loss: 0.0408 - acc: 0.9876 - val_loss: 3.1702 - val_acc: 0.6942
Epoch 00066: val_acc did not improve from 0.70677
Epoch 67/100
- 532s - loss: 0.0520 - acc: 0.9853 - val_loss: 3.2619 - val_acc: 0.6988
Epoch 00067: val_acc did not improve from 0.70677
Epoch 68/100
- 540s - loss: 0.0318 - acc: 0.9890 - val_loss: 3.1641 - val_acc: 0.7160
Epoch 00068: val_acc improved from 0.70677 to 0.71596, saving model to weights\weights_RNN_0.hdf5
Epoch 69/100
- 535s - loss: 0.0310 - acc: 0.9909 - val_loss: 3.5259 - val_acc: 0.6913
Epoch 00069: val_acc did not improve from 0.71596
Epoch 70/100
- 541s - loss: 0.0418 - acc: 0.9865 - val_loss: 3.9242 - val_acc: 0.6374
Epoch 00070: val_acc did not improve from 0.71596
Epoch 71/100
- 531s - loss: 0.0414 - acc: 0.9877 - val_loss: 3.9219 - val_acc: 0.6537
Epoch 00071: val_acc did not improve from 0.71596
Epoch 72/100
- 525s - loss: 0.0418 - acc: 0.9897 - val_loss: 3.6954 - val_acc: 0.6504
Epoch 00072: val_acc did not improve from 0.71596
Epoch 73/100
- 539s - loss: 0.0329 - acc: 0.9896 - val_loss: 3.5067 - val_acc: 0.6800
Epoch 00073: val_acc did not improve from 0.71596
Epoch 74/100
- 563s - loss: 0.0407 - acc: 0.9893 - val_loss: 3.5700 - val_acc: 0.6729
Epoch 00074: val_acc did not improve from 0.71596
Epoch 75/100
- 535s - loss: 0.0301 - acc: 0.9901 - val_loss: 3.9602 - val_acc: 0.6366
Epoch 00075: val_acc did not improve from 0.71596
Epoch 76/100
- 532s - loss: 0.0363 - acc: 0.9885 - val_loss: 3.5299 - val_acc: 0.7001
Epoch 00076: val_acc did not improve from 0.71596
Epoch 77/100
- 548s - loss: 0.0348 - acc: 0.9890 - val_loss: 3.4072 - val_acc: 0.7013
Epoch 00077: val_acc did not improve from 0.71596
Epoch 78/100
- 569s - loss: 0.0403 - acc: 0.9890 - val_loss: 3.6074 - val_acc: 0.6637
Epoch 00078: val_acc did not improve from 0.71596
Epoch 79/100
- 569s - loss: 0.0490 - acc: 0.9873 - val_loss: 3.3901 - val_acc: 0.6934
Epoch 00079: val_acc did not improve from 0.71596
Epoch 80/100
- 557s - loss: 0.0386 - acc: 0.9891 - val_loss: 3.3877 - val_acc: 0.7072
Epoch 00080: val_acc did not improve from 0.71596
Epoch 81/100
- 505s - loss: 0.0350 - acc: 0.9903 - val_loss: 3.5415 - val_acc: 0.7068
Epoch 00081: val_acc did not improve from 0.71596
Epoch 82/100
- 535s - loss: 0.0420 - acc: 0.9891 - val_loss: 3.4312 - val_acc: 0.7043
Epoch 00082: val_acc did not improve from 0.71596
Epoch 83/100
- 535s - loss: 0.0322 - acc: 0.9904 - val_loss: 3.5121 - val_acc: 0.6967
Epoch 00083: val_acc did not improve from 0.71596
Epoch 84/100
- 524s - loss: 0.0362 - acc: 0.9892 - val_loss: 3.6263 - val_acc: 0.6855
Epoch 00084: val_acc did not improve from 0.71596
Epoch 85/100
- 530s - loss: 0.0249 - acc: 0.9910 - val_loss: 3.6276 - val_acc: 0.6855
Epoch 00085: val_acc did not improve from 0.71596
Epoch 86/100
- 520s - loss: 0.0296 - acc: 0.9917 - val_loss: 3.5167 - val_acc: 0.6984
Epoch 00086: val_acc did not improve from 0.71596
Epoch 87/100
- 523s - loss: 0.0419 - acc: 0.9883 - val_loss: 3.7824 - val_acc: 0.6779
Epoch 00087: val_acc did not improve from 0.71596
Epoch 88/100
- 527s - loss: 0.0227 - acc: 0.9933 - val_loss: 3.4899 - val_acc: 0.7013
Epoch 00088: val_acc did not improve from 0.71596
Epoch 89/100
- 531s - loss: 0.0383 - acc: 0.9894 - val_loss: 3.6205 - val_acc: 0.6834
Epoch 00089: val_acc did not improve from 0.71596
Epoch 90/100
- 533s - loss: 0.0337 - acc: 0.9919 - val_loss: 3.6725 - val_acc: 0.6863
Epoch 00090: val_acc did not improve from 0.71596
Epoch 91/100
- 822s - loss: 0.0365 - acc: 0.9902 - val_loss: 3.3955 - val_acc: 0.7005
Epoch 00091: val_acc did not improve from 0.71596
Epoch 92/100
- 1069s - loss: 0.0321 - acc: 0.9904 - val_loss: 3.3287 - val_acc: 0.7164
Epoch 00092: val_acc improved from 0.71596 to 0.71637, saving model to weights\weights_RNN_0.hdf5
Epoch 93/100
- 903s - loss: 0.0293 - acc: 0.9915 - val_loss: 3.5323 - val_acc: 0.6876
Epoch 00093: val_acc did not improve from 0.71637
Epoch 94/100
- 890s - loss: 0.0263 - acc: 0.9906 - val_loss: 3.4182 - val_acc: 0.7063
Epoch 00094: val_acc did not improve from 0.71637
Epoch 95/100
- 808s - loss: 0.0239 - acc: 0.9925 - val_loss: 3.7548 - val_acc: 0.6871
Epoch 00095: val_acc did not improve from 0.71637
Epoch 96/100
- 814s - loss: 0.0240 - acc: 0.9920 - val_loss: 3.4282 - val_acc: 0.7034
Epoch 00096: val_acc did not improve from 0.71637
Epoch 97/100
- 832s - loss: 0.0240 - acc: 0.9930 - val_loss: 3.4407 - val_acc: 0.7114
Epoch 00097: val_acc did not improve from 0.71637
Epoch 98/100
- 856s - loss: 0.0373 - acc: 0.9908 - val_loss: 3.7107 - val_acc: 0.6884
Epoch 00098: val_acc did not improve from 0.71637
Epoch 99/100
- 749s - loss: 0.0255 - acc: 0.9920 - val_loss: 3.5944 - val_acc: 0.6972
Epoch 00099: val_acc did not improve from 0.71637
Epoch 100/100
- 938s - loss: 0.0237 - acc: 0.9928 - val_loss: 3.5825 - val_acc: 0.7013
Epoch 00100: val_acc did not improve from 0.71637
RNN 1
2
99
<keras.optimizers.RMSprop object at 0x7ff0e07f6208>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 602s - loss: 3.1192 - acc: 0.0913 - val_loss: 2.6589 - val_acc: 0.1558
Epoch 00001: val_acc improved from -inf to 0.15581, saving model to weights\weights_RNN_1.hdf5
Epoch 2/100
- 639s - loss: 2.5080 - acc: 0.1824 - val_loss: 2.2314 - val_acc: 0.2485
Epoch 00002: val_acc improved from 0.15581 to 0.24854, saving model to weights\weights_RNN_1.hdf5
Epoch 3/100
- 572s - loss: 2.1150 - acc: 0.2864 - val_loss: 1.8693 - val_acc: 0.3776
Epoch 00003: val_acc improved from 0.24854 to 0.37761, saving model to weights\weights_RNN_1.hdf5
Epoch 4/100
- 612s - loss: 1.7069 - acc: 0.4343 - val_loss: 1.5585 - val_acc: 0.4833
Epoch 00004: val_acc improved from 0.37761 to 0.48329, saving model to weights\weights_RNN_1.hdf5
Epoch 5/100
- 627s - loss: 1.3073 - acc: 0.5617 - val_loss: 1.2494 - val_acc: 0.6140
Epoch 00005: val_acc improved from 0.48329 to 0.61404, saving model to weights\weights_RNN_1.hdf5
Epoch 6/100
- 688s - loss: 0.9703 - acc: 0.6866 - val_loss: 0.9871 - val_acc: 0.7047
Epoch 00006: val_acc improved from 0.61404 to 0.70468, saving model to weights\weights_RNN_1.hdf5
Epoch 7/100
- 691s - loss: 0.7670 - acc: 0.7608 - val_loss: 0.8767 - val_acc: 0.7623
Epoch 00007: val_acc improved from 0.70468 to 0.76232, saving model to weights\weights_RNN_1.hdf5
Epoch 8/100
- 660s - loss: 0.6090 - acc: 0.8146 - val_loss: 0.8632 - val_acc: 0.7744
Epoch 00008: val_acc improved from 0.76232 to 0.77444, saving model to weights\weights_RNN_1.hdf5
Epoch 9/100
- 688s - loss: 0.4891 - acc: 0.8539 - val_loss: 0.8302 - val_acc: 0.7916
Epoch 00009: val_acc improved from 0.77444 to 0.79156, saving model to weights\weights_RNN_1.hdf5
Epoch 10/100
- 724s - loss: 0.4015 - acc: 0.8809 - val_loss: 0.8235 - val_acc: 0.8049
Epoch 00010: val_acc improved from 0.79156 to 0.80493, saving model to weights\weights_RNN_1.hdf5
Epoch 11/100
- 697s - loss: 0.3249 - acc: 0.9077 - val_loss: 0.9020 - val_acc: 0.8012
Epoch 00011: val_acc did not improve from 0.80493
Epoch 12/100
- 694s - loss: 0.2727 - acc: 0.9240 - val_loss: 0.9440 - val_acc: 0.8083
Epoch 00012: val_acc improved from 0.80493 to 0.80827, saving model to weights\weights_RNN_1.hdf5
Epoch 13/100
- 640s - loss: 0.2220 - acc: 0.9370 - val_loss: 0.9457 - val_acc: 0.8062
Epoch 00013: val_acc did not improve from 0.80827
Epoch 14/100
- 557s - loss: 0.1903 - acc: 0.9451 - val_loss: 1.0360 - val_acc: 0.8179
Epoch 00014: val_acc improved from 0.80827 to 0.81788, saving model to weights\weights_RNN_1.hdf5
Epoch 15/100
- 535s - loss: 0.1495 - acc: 0.9571 - val_loss: 0.9993 - val_acc: 0.8208
Epoch 00015: val_acc improved from 0.81788 to 0.82080, saving model to weights\weights_RNN_1.hdf5
Epoch 16/100
- 565s - loss: 0.1248 - acc: 0.9647 - val_loss: 1.0833 - val_acc: 0.8166
...
Epoch 00099: val_acc did not improve from 0.82247
Epoch 100/100
- 415s - loss: 0.0071 - acc: 0.9984 - val_loss: 2.4204 - val_acc: 0.8141
Epoch 00100: val_acc did not improve from 0.82247
RNN 2
3
119
<keras.optimizers.Adam object at 0x7ff0df2da4a8>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 562s - loss: 3.1450 - acc: 0.0820 - val_loss: 2.6098 - val_acc: 0.1433
Epoch 00001: val_acc improved from -inf to 0.14327, saving model to weights\weights_RNN_2.hdf5
Epoch 2/100
- 562s - loss: 2.3437 - acc: 0.2197 - val_loss: 1.9166 - val_acc: 0.3417
Epoch 00002: val_acc improved from 0.14327 to 0.34169, saving model to weights\weights_RNN_2.hdf5
Epoch 3/100
- 558s - loss: 1.6448 - acc: 0.4426 - val_loss: 1.3043 - val_acc: 0.5622
Epoch 00003: val_acc improved from 0.34169 to 0.56224, saving model to weights\weights_RNN_2.hdf5
Epoch 4/100
- 569s - loss: 1.0390 - acc: 0.6541 - val_loss: 0.9435 - val_acc: 0.7026
Epoch 00004: val_acc improved from 0.56224 to 0.70259, saving model to weights\weights_RNN_2.hdf5
Epoch 5/100
- 561s - loss: 0.6602 - acc: 0.7848 - val_loss: 0.9575 - val_acc: 0.7352
Epoch 00005: val_acc improved from 0.70259 to 0.73517, saving model to weights\weights_RNN_2.hdf5
Epoch 6/100
- 565s - loss: 0.4394 - acc: 0.8666 - val_loss: 0.9166 - val_acc: 0.7707
Epoch 00006: val_acc improved from 0.73517 to 0.77068, saving model to weights\weights_RNN_2.hdf5
Epoch 7/100
- 563s - loss: 0.3009 - acc: 0.9101 - val_loss: 0.9634 - val_acc: 0.7736
Epoch 00007: val_acc improved from 0.77068 to 0.77360, saving model to weights\weights_RNN_2.hdf5
Epoch 8/100
- 565s - loss: 0.2182 - acc: 0.9361 - val_loss: 1.1147 - val_acc: 0.7719
Epoch 00008: val_acc did not improve from 0.77360
Epoch 9/100
- 564s - loss: 0.1498 - acc: 0.9569 - val_loss: 1.1924 - val_acc: 0.7778
Epoch 00009: val_acc improved from 0.77360 to 0.77778, saving model to weights\weights_RNN_2.hdf5
Epoch 10/100
- 564s - loss: 0.1104 - acc: 0.9681 - val_loss: 1.2466 - val_acc: 0.7778
Epoch 00010: val_acc improved from 0.77778 to 0.77778, saving model to weights\weights_RNN_2.hdf5
Epoch 11/100
- 565s - loss: 0.0851 - acc: 0.9770 - val_loss: 1.3057 - val_acc: 0.7707
Epoch 00011: val_acc did not improve from 0.77778
Epoch 12/100
- 565s - loss: 0.0700 - acc: 0.9806 - val_loss: 1.3911 - val_acc: 0.7778
Epoch 00012: val_acc did not improve from 0.77778
Epoch 13/100
- 562s - loss: 0.0722 - acc: 0.9796 - val_loss: 1.4207 - val_acc: 0.7774
Epoch 00013: val_acc did not improve from 0.77778
Epoch 14/100
- 564s - loss: 0.0628 - acc: 0.9814 - val_loss: 1.4512 - val_acc: 0.7665
Epoch 00014: val_acc did not improve from 0.77778
Epoch 15/100
- 559s - loss: 0.0486 - acc: 0.9876 - val_loss: 1.4833 - val_acc: 0.7782
Epoch 00015: val_acc improved from 0.77778 to 0.77820, saving model to weights\weights_RNN_2.hdf5
Epoch 16/100
- 562s - loss: 0.0532 - acc: 0.9845 - val_loss: 1.5212 - val_acc: 0.7749
Epoch 00016: val_acc did not improve from 0.77820
Epoch 17/100
- 562s - loss: 0.0352 - acc: 0.9893 - val_loss: 1.7025 - val_acc: 0.7690
Epoch 00017: val_acc did not improve from 0.77820
Epoch 18/100
- 567s - loss: 0.0393 - acc: 0.9896 - val_loss: 1.6166 - val_acc: 0.7602
Epoch 00018: val_acc did not improve from 0.77820
Epoch 19/100
- 559s - loss: 0.0415 - acc: 0.9885 - val_loss: 1.6914 - val_acc: 0.7732
Epoch 00019: val_acc did not improve from 0.77820
Epoch 20/100
- 590s - loss: 0.0413 - acc: 0.9880 - val_loss: 1.5904 - val_acc: 0.7832
Epoch 00020: val_acc improved from 0.77820 to 0.78321, saving model to weights\weights_RNN_2.hdf5
Epoch 21/100
- 577s - loss: 0.0492 - acc: 0.9862 - val_loss: 1.7063 - val_acc: 0.7749
Epoch 00021: val_acc did not improve from 0.78321
Epoch 22/100
- 581s - loss: 0.0500 - acc: 0.9863 - val_loss: 1.6024 - val_acc: 0.7782
Epoch 00022: val_acc did not improve from 0.78321
Epoch 23/100
- 582s - loss: 0.0384 - acc: 0.9900 - val_loss: 1.5958 - val_acc: 0.7895
Epoch 00023: val_acc improved from 0.78321 to 0.78947, saving model to weights\weights_RNN_2.hdf5
Epoch 24/100
- 564s - loss: 0.0202 - acc: 0.9939 - val_loss: 1.7483 - val_acc: 0.7861
Epoch 00024: val_acc did not improve from 0.78947
Epoch 25/100
- 559s - loss: 0.0243 - acc: 0.9931 - val_loss: 1.7565 - val_acc: 0.7786
Epoch 00025: val_acc did not improve from 0.78947
Epoch 26/100
- 570s - loss: 0.0303 - acc: 0.9915 - val_loss: 1.6901 - val_acc: 0.7870
Epoch 00026: val_acc did not improve from 0.78947
Epoch 27/100
- 591s - loss: 0.0202 - acc: 0.9947 - val_loss: 1.8889 - val_acc: 0.7774
Epoch 00027: val_acc did not improve from 0.78947
Epoch 28/100
- 596s - loss: 0.0274 - acc: 0.9927 - val_loss: 1.7228 - val_acc: 0.7920
Epoch 00028: val_acc improved from 0.78947 to 0.79198, saving model to weights\weights_RNN_2.hdf5
Epoch 29/100
- 592s - loss: 0.0267 - acc: 0.9928 - val_loss: 1.7685 - val_acc: 0.7765
Epoch 00029: val_acc did not improve from 0.79198
Epoch 30/100
- 644s - loss: 0.0423 - acc: 0.9883 - val_loss: 1.8158 - val_acc: 0.7828
Epoch 00030: val_acc did not improve from 0.79198
Epoch 31/100
- 650s - loss: 0.0299 - acc: 0.9924 - val_loss: 1.8958 - val_acc: 0.7740
Epoch 00031: val_acc did not improve from 0.79198
Epoch 32/100
- 650s - loss: 0.0276 - acc: 0.9922 - val_loss: 1.7903 - val_acc: 0.7899
Epoch 00032: val_acc did not improve from 0.79198
Epoch 33/100
- 633s - loss: 0.0239 - acc: 0.9927 - val_loss: 1.9673 - val_acc: 0.7690
Epoch 00033: val_acc did not improve from 0.79198
Epoch 34/100
- 612s - loss: 0.0322 - acc: 0.9916 - val_loss: 1.9237 - val_acc: 0.7774
Epoch 00034: val_acc did not improve from 0.79198
Epoch 35/100
- 603s - loss: 0.0380 - acc: 0.9912 - val_loss: 1.9174 - val_acc: 0.7723
Epoch 00035: val_acc did not improve from 0.79198
Epoch 36/100
- 577s - loss: 0.0210 - acc: 0.9947 - val_loss: 1.6420 - val_acc: 0.8028
Epoch 00036: val_acc improved from 0.79198 to 0.80284, saving model to weights\weights_RNN_2.hdf5
Epoch 37/100
- 572s - loss: 0.0165 - acc: 0.9956 - val_loss: 1.7218 - val_acc: 0.7949
Epoch 00037: val_acc did not improve from 0.80284
Epoch 38/100
- 561s - loss: 0.0095 - acc: 0.9960 - val_loss: 1.8369 - val_acc: 0.7882
Epoch 00038: val_acc did not improve from 0.80284
Epoch 39/100
- 562s - loss: 0.0152 - acc: 0.9956 - val_loss: 1.9850 - val_acc: 0.7765
Epoch 00039: val_acc did not improve from 0.80284
Epoch 40/100
- 568s - loss: 0.0206 - acc: 0.9946 - val_loss: 1.9381 - val_acc: 0.7899
Epoch 00040: val_acc did not improve from 0.80284
Epoch 41/100
- 562s - loss: 0.0224 - acc: 0.9944 - val_loss: 1.9268 - val_acc: 0.7874
Epoch 00041: val_acc did not improve from 0.80284
Epoch 42/100
- 541s - loss: 0.0214 - acc: 0.9947 - val_loss: 1.9528 - val_acc: 0.7891
Epoch 00042: val_acc did not improve from 0.80284
Epoch 43/100
- 543s - loss: 0.0287 - acc: 0.9921 - val_loss: 2.0765 - val_acc: 0.7728
Epoch 00043: val_acc did not improve from 0.80284
Epoch 44/100
- 549s - loss: 0.0282 - acc: 0.9924 - val_loss: 2.0929 - val_acc: 0.7828
Epoch 00044: val_acc did not improve from 0.80284
Epoch 45/100
- 553s - loss: 0.0345 - acc: 0.9910 - val_loss: 1.8607 - val_acc: 0.7861
Epoch 00045: val_acc did not improve from 0.80284
Epoch 46/100
- 553s - loss: 0.0347 - acc: 0.9909 - val_loss: 2.0510 - val_acc: 0.7774
Epoch 00046: val_acc did not improve from 0.80284
Epoch 47/100
- 575s - loss: 0.0275 - acc: 0.9924 - val_loss: 1.8327 - val_acc: 0.7974
Epoch 00047: val_acc did not improve from 0.80284
Epoch 48/100
- 583s - loss: 0.0160 - acc: 0.9957 - val_loss: 1.9240 - val_acc: 0.7916
Epoch 00048: val_acc did not improve from 0.80284
Epoch 49/100
- 621s - loss: 0.0059 - acc: 0.9979 - val_loss: 1.8761 - val_acc: 0.8066
Epoch 00049: val_acc improved from 0.80284 to 0.80660, saving model to weights\weights_RNN_2.hdf5
Epoch 50/100
- 633s - loss: 0.0095 - acc: 0.9978 - val_loss: 1.9150 - val_acc: 0.7991
...
Epoch 00099: val_acc did not improve from 0.80911
Epoch 100/100
- 526s - loss: 0.0136 - acc: 0.9972 - val_loss: 2.3657 - val_acc: 0.7765
Epoch 00100: val_acc did not improve from 0.80911
CNN 0
Filter 5
Node 465
<keras.optimizers.Adam object at 0x7ff0dddad668>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 35s - loss: 3.5457 - acc: 0.0344 - val_loss: 3.3673 - val_acc: 0.0359
Epoch 00001: val_acc improved from -inf to 0.03592, saving model to weights\weights_CNN_0.hdf5
Epoch 2/100
- 16s - loss: 3.1529 - acc: 0.0828 - val_loss: 3.0458 - val_acc: 0.1157
Epoch 00002: val_acc improved from 0.03592 to 0.11571, saving model to weights\weights_CNN_0.hdf5
Epoch 3/100
- 16s - loss: 2.6238 - acc: 0.1558 - val_loss: 2.7819 - val_acc: 0.1596
Epoch 00003: val_acc improved from 0.11571 to 0.15957, saving model to weights\weights_CNN_0.hdf5
Epoch 4/100
- 16s - loss: 2.3458 - acc: 0.2077 - val_loss: 2.5020 - val_acc: 0.2222
Epoch 00004: val_acc improved from 0.15957 to 0.22222, saving model to weights\weights_CNN_0.hdf5
Epoch 5/100
- 16s - loss: 2.0654 - acc: 0.2936 - val_loss: 2.2700 - val_acc: 0.2794
Epoch 00005: val_acc improved from 0.22222 to 0.27945, saving model to weights\weights_CNN_0.hdf5
Epoch 6/100
- 16s - loss: 1.7404 - acc: 0.3902 - val_loss: 2.0848 - val_acc: 0.3655
Epoch 00006: val_acc improved from 0.27945 to 0.36550, saving model to weights\weights_CNN_0.hdf5
Epoch 7/100
- 16s - loss: 1.3925 - acc: 0.5163 - val_loss: 1.6499 - val_acc: 0.5230
Epoch 00007: val_acc improved from 0.36550 to 0.52297, saving model to weights\weights_CNN_0.hdf5
Epoch 8/100
- 16s - loss: 1.1197 - acc: 0.6129 - val_loss: 1.6440 - val_acc: 0.5000
Epoch 00008: val_acc did not improve from 0.52297
Epoch 9/100
- 16s - loss: 0.9011 - acc: 0.6941 - val_loss: 1.2881 - val_acc: 0.6241
Epoch 00009: val_acc improved from 0.52297 to 0.62406, saving model to weights\weights_CNN_0.hdf5
Epoch 10/100
- 16s - loss: 0.7151 - acc: 0.7647 - val_loss: 1.2029 - val_acc: 0.6591
Epoch 00010: val_acc improved from 0.62406 to 0.65915, saving model to weights\weights_CNN_0.hdf5
Epoch 11/100
- 16s - loss: 0.5664 - acc: 0.8132 - val_loss: 1.1157 - val_acc: 0.6859
Epoch 00011: val_acc improved from 0.65915 to 0.68588, saving model to weights\weights_CNN_0.hdf5
Epoch 12/100
- 15s - loss: 0.4694 - acc: 0.8533 - val_loss: 1.0223 - val_acc: 0.7105
Epoch 00012: val_acc improved from 0.68588 to 0.71053, saving model to weights\weights_CNN_0.hdf5
Epoch 13/100
- 16s - loss: 0.3707 - acc: 0.8829 - val_loss: 0.9569 - val_acc: 0.7335
Epoch 00013: val_acc improved from 0.71053 to 0.73350, saving model to weights\weights_CNN_0.hdf5
Epoch 14/100
- 16s - loss: 0.3311 - acc: 0.8979 - val_loss: 0.9396 - val_acc: 0.7435
Epoch 00014: val_acc improved from 0.73350 to 0.74353, saving model to weights\weights_CNN_0.hdf5
Epoch 15/100
- 16s - loss: 0.2643 - acc: 0.9177 - val_loss: 1.0162 - val_acc: 0.7239
Epoch 00015: val_acc did not improve from 0.74353
Epoch 16/100
- 16s - loss: 0.2342 - acc: 0.9303 - val_loss: 0.8967 - val_acc: 0.7586
Epoch 00016: val_acc improved from 0.74353 to 0.75856, saving model to weights\weights_CNN_0.hdf5
Epoch 17/100
- 16s - loss: 0.2096 - acc: 0.9368 - val_loss: 0.9957 - val_acc: 0.7331
Epoch 00017: val_acc did not improve from 0.75856
Epoch 18/100
- 15s - loss: 0.2007 - acc: 0.9394 - val_loss: 0.9404 - val_acc: 0.7548
Epoch 00018: val_acc did not improve from 0.75856
Epoch 19/100
- 16s - loss: 0.1669 - acc: 0.9517 - val_loss: 0.9567 - val_acc: 0.7569
Epoch 00019: val_acc did not improve from 0.75856
Epoch 20/100
- 15s - loss: 0.1573 - acc: 0.9547 - val_loss: 0.9228 - val_acc: 0.7636
Epoch 00020: val_acc improved from 0.75856 to 0.76358, saving model to weights\weights_CNN_0.hdf5
Epoch 21/100
- 15s - loss: 0.1598 - acc: 0.9528 - val_loss: 0.9539 - val_acc: 0.7623
Epoch 00021: val_acc did not improve from 0.76358
Epoch 22/100
- 16s - loss: 0.1324 - acc: 0.9626 - val_loss: 0.9476 - val_acc: 0.7765
Epoch 00022: val_acc improved from 0.76358 to 0.77652, saving model to weights\weights_CNN_0.hdf5
Epoch 23/100
- 16s - loss: 0.1515 - acc: 0.9579 - val_loss: 0.9376 - val_acc: 0.7794
Epoch 00023: val_acc improved from 0.77652 to 0.77945, saving model to weights\weights_CNN_0.hdf5
Epoch 24/100
- 15s - loss: 0.1258 - acc: 0.9654 - val_loss: 0.9445 - val_acc: 0.7707
Epoch 00024: val_acc did not improve from 0.77945
Epoch 25/100
- 16s - loss: 0.1328 - acc: 0.9623 - val_loss: 0.9399 - val_acc: 0.7774
Epoch 00025: val_acc did not improve from 0.77945
Epoch 26/100
- 16s - loss: 0.1365 - acc: 0.9658 - val_loss: 0.9852 - val_acc: 0.7761
Epoch 00026: val_acc did not improve from 0.77945
Epoch 27/100
- 16s - loss: 0.1161 - acc: 0.9699 - val_loss: 1.0045 - val_acc: 0.7765
Epoch 00027: val_acc did not improve from 0.77945
Epoch 28/100
- 15s - loss: 0.1030 - acc: 0.9748 - val_loss: 0.9620 - val_acc: 0.7874
Epoch 00028: val_acc improved from 0.77945 to 0.78739, saving model to weights\weights_CNN_0.hdf5
Epoch 29/100
- 15s - loss: 0.1223 - acc: 0.9669 - val_loss: 1.0282 - val_acc: 0.7678
Epoch 00029: val_acc did not improve from 0.78739
Epoch 30/100
- 15s - loss: 0.1395 - acc: 0.9679 - val_loss: 0.9603 - val_acc: 0.7949
Epoch 00030: val_acc improved from 0.78739 to 0.79490, saving model to weights\weights_CNN_0.hdf5
Epoch 31/100
- 16s - loss: 0.1292 - acc: 0.9704 - val_loss: 0.9944 - val_acc: 0.7794
Epoch 00031: val_acc did not improve from 0.79490
Epoch 32/100
- 15s - loss: 0.1228 - acc: 0.9703 - val_loss: 0.9323 - val_acc: 0.7949
Epoch 00032: val_acc did not improve from 0.79490
Epoch 33/100
- 15s - loss: 0.1089 - acc: 0.9743 - val_loss: 0.9966 - val_acc: 0.7891
Epoch 00033: val_acc did not improve from 0.79490
Epoch 34/100
- 16s - loss: 0.0961 - acc: 0.9762 - val_loss: 0.9813 - val_acc: 0.7991
Epoch 00034: val_acc improved from 0.79490 to 0.79908, saving model to weights\weights_CNN_0.hdf5
Epoch 35/100
- 16s - loss: 0.1283 - acc: 0.9748 - val_loss: 1.1216 - val_acc: 0.7828
Epoch 00035: val_acc did not improve from 0.79908
Epoch 36/100
- 15s - loss: 0.1122 - acc: 0.9760 - val_loss: 1.1793 - val_acc: 0.7774
Epoch 00036: val_acc did not improve from 0.79908
Epoch 37/100
- 16s - loss: 0.1766 - acc: 0.9636 - val_loss: 0.9285 - val_acc: 0.8108
Epoch 00037: val_acc improved from 0.79908 to 0.81078, saving model to weights\weights_CNN_0.hdf5
Epoch 38/100
- 16s - loss: 0.1106 - acc: 0.9757 - val_loss: 1.0482 - val_acc: 0.7903
Epoch 00038: val_acc did not improve from 0.81078
Epoch 39/100
- 15s - loss: 0.1110 - acc: 0.9762 - val_loss: 1.0710 - val_acc: 0.7999
Epoch 00039: val_acc did not improve from 0.81078
Epoch 40/100
- 16s - loss: 0.1319 - acc: 0.9743 - val_loss: 1.0491 - val_acc: 0.8033
Epoch 00040: val_acc did not improve from 0.81078
Epoch 41/100
- 16s - loss: 0.1066 - acc: 0.9773 - val_loss: 1.1048 - val_acc: 0.7978
Epoch 00041: val_acc did not improve from 0.81078
Epoch 42/100
- 15s - loss: 0.0913 - acc: 0.9823 - val_loss: 1.1703 - val_acc: 0.7891
Epoch 00042: val_acc did not improve from 0.81078
Epoch 43/100
- 15s - loss: 0.1133 - acc: 0.9786 - val_loss: 1.1666 - val_acc: 0.7974
Epoch 00043: val_acc did not improve from 0.81078
Epoch 44/100
- 15s - loss: 0.1278 - acc: 0.9739 - val_loss: 1.0674 - val_acc: 0.8012
Epoch 00044: val_acc did not improve from 0.81078
Epoch 45/100
- 16s - loss: 0.0951 - acc: 0.9812 - val_loss: 1.1186 - val_acc: 0.8008
Epoch 00045: val_acc did not improve from 0.81078
Epoch 46/100
- 15s - loss: 0.0926 - acc: 0.9816 - val_loss: 1.0971 - val_acc: 0.8091
Epoch 00046: val_acc did not improve from 0.81078
Epoch 47/100
- 15s - loss: 0.0850 - acc: 0.9829 - val_loss: 1.0583 - val_acc: 0.8045
Epoch 00047: val_acc did not improve from 0.81078
Epoch 48/100
- 16s - loss: 0.0795 - acc: 0.9846 - val_loss: 1.2156 - val_acc: 0.8049
Epoch 00048: val_acc did not improve from 0.81078
Epoch 49/100
- 16s - loss: 0.1116 - acc: 0.9811 - val_loss: 1.1635 - val_acc: 0.7953
Epoch 00049: val_acc did not improve from 0.81078
Epoch 50/100
- 15s - loss: 0.1279 - acc: 0.9797 - val_loss: 1.2867 - val_acc: 0.8008
Epoch 00050: val_acc did not improve from 0.81078
Epoch 51/100
- 16s - loss: 0.1513 - acc: 0.9738 - val_loss: 1.1304 - val_acc: 0.8066
Epoch 00051: val_acc did not improve from 0.81078
Epoch 52/100
- 16s - loss: 0.1244 - acc: 0.9808 - val_loss: 1.2704 - val_acc: 0.7962
Epoch 00052: val_acc did not improve from 0.81078
Epoch 53/100
- 16s - loss: 0.1870 - acc: 0.9725 - val_loss: 1.0882 - val_acc: 0.8133
Epoch 00053: val_acc improved from 0.81078 to 0.81328, saving model to weights\weights_CNN_0.hdf5
Epoch 54/100
- 16s - loss: 0.2051 - acc: 0.9694 - val_loss: 1.3659 - val_acc: 0.7991
Epoch 00054: val_acc did not improve from 0.81328
Epoch 55/100
- 15s - loss: 0.2977 - acc: 0.9630 - val_loss: 1.2798 - val_acc: 0.8204
Epoch 00055: val_acc improved from 0.81328 to 0.82038, saving model to weights\weights_CNN_0.hdf5
Epoch 56/100
- 15s - loss: 0.1299 - acc: 0.9790 - val_loss: 1.2250 - val_acc: 0.8137
...
Epoch 99/100
- 15s - loss: 0.4090 - acc: 0.9697 - val_loss: 2.2313 - val_acc: 0.8266
Epoch 00099: val_acc did not improve from 0.83835
Epoch 100/100
- 15s - loss: 0.3692 - acc: 0.9726 - val_loss: 2.3687 - val_acc: 0.8275
Epoch 00100: val_acc did not improve from 0.83835
CNN 1
Filter 6
Node 347
<keras.optimizers.Adam object at 0x7ff0da830240>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 25s - loss: 3.5243 - acc: 0.0389 - val_loss: 3.3594 - val_acc: 0.0689
Epoch 00001: val_acc improved from -inf to 0.06892, saving model to weights\weights_CNN_1.hdf5
Epoch 2/100
- 13s - loss: 2.7901 - acc: 0.1331 - val_loss: 2.6931 - val_acc: 0.2343
Epoch 00002: val_acc improved from 0.06892 to 0.23434, saving model to weights\weights_CNN_1.hdf5
Epoch 3/100
- 13s - loss: 2.0914 - acc: 0.2692 - val_loss: 2.1964 - val_acc: 0.4089
Epoch 00003: val_acc improved from 0.23434 to 0.40894, saving model to weights\weights_CNN_1.hdf5
Epoch 4/100
- 13s - loss: 1.5188 - acc: 0.4642 - val_loss: 1.6398 - val_acc: 0.5188
Epoch 00004: val_acc improved from 0.40894 to 0.51880, saving model to weights\weights_CNN_1.hdf5
Epoch 5/100
- 14s - loss: 1.0329 - acc: 0.6438 - val_loss: 1.1347 - val_acc: 0.7126
Epoch 00005: val_acc improved from 0.51880 to 0.71261, saving model to weights\weights_CNN_1.hdf5
Epoch 6/100
- 13s - loss: 0.7135 - acc: 0.7649 - val_loss: 0.9307 - val_acc: 0.7619
Epoch 00006: val_acc improved from 0.71261 to 0.76190, saving model to weights\weights_CNN_1.hdf5
Epoch 7/100
- 14s - loss: 0.5280 - acc: 0.8336 - val_loss: 0.8341 - val_acc: 0.7895
Epoch 00007: val_acc improved from 0.76190 to 0.78947, saving model to weights\weights_CNN_1.hdf5
Epoch 8/100
- 14s - loss: 0.3857 - acc: 0.8822 - val_loss: 0.7689 - val_acc: 0.7941
Epoch 00008: val_acc improved from 0.78947 to 0.79407, saving model to weights\weights_CNN_1.hdf5
Epoch 9/100
- 15s - loss: 0.3073 - acc: 0.9123 - val_loss: 0.7414 - val_acc: 0.8108
Epoch 00009: val_acc improved from 0.79407 to 0.81078, saving model to weights\weights_CNN_1.hdf5
Epoch 10/100
- 13s - loss: 0.2472 - acc: 0.9313 - val_loss: 0.7146 - val_acc: 0.8079
Epoch 00010: val_acc did not improve from 0.81078
Epoch 11/100
- 13s - loss: 0.2178 - acc: 0.9407 - val_loss: 0.7507 - val_acc: 0.8116
Epoch 00011: val_acc improved from 0.81078 to 0.81161, saving model to weights\weights_CNN_1.hdf5
Epoch 12/100
- 13s - loss: 0.1551 - acc: 0.9554 - val_loss: 0.7309 - val_acc: 0.8200
Epoch 00012: val_acc improved from 0.81161 to 0.81997, saving model to weights\weights_CNN_1.hdf5
Epoch 13/100
- 13s - loss: 0.1500 - acc: 0.9584 - val_loss: 0.7546 - val_acc: 0.8141
Epoch 00013: val_acc did not improve from 0.81997
Epoch 14/100
- 13s - loss: 0.1642 - acc: 0.9568 - val_loss: 0.7293 - val_acc: 0.8237
Epoch 00014: val_acc improved from 0.81997 to 0.82373, saving model to weights\weights_CNN_1.hdf5
Epoch 15/100
- 13s - loss: 0.1270 - acc: 0.9665 - val_loss: 0.6991 - val_acc: 0.8329
Epoch 00015: val_acc improved from 0.82373 to 0.83292, saving model to weights\weights_CNN_1.hdf5
Epoch 16/100
- 13s - loss: 0.0976 - acc: 0.9742 - val_loss: 0.7409 - val_acc: 0.8212
Epoch 00016: val_acc did not improve from 0.83292
Epoch 17/100
- 13s - loss: 0.1149 - acc: 0.9693 - val_loss: 0.7916 - val_acc: 0.8246
Epoch 00017: val_acc did not improve from 0.83292
Epoch 18/100
- 13s - loss: 0.1047 - acc: 0.9749 - val_loss: 0.7482 - val_acc: 0.8254
Epoch 00018: val_acc did not improve from 0.83292
Epoch 19/100
- 13s - loss: 0.1123 - acc: 0.9744 - val_loss: 0.7892 - val_acc: 0.8304
Epoch 00019: val_acc did not improve from 0.83292
Epoch 20/100
- 13s - loss: 0.1329 - acc: 0.9690 - val_loss: 0.8295 - val_acc: 0.8187
Epoch 00020: val_acc did not improve from 0.83292
Epoch 21/100
- 13s - loss: 0.1415 - acc: 0.9669 - val_loss: 0.8018 - val_acc: 0.8124
Epoch 00021: val_acc did not improve from 0.83292
Epoch 22/100
- 13s - loss: 0.0893 - acc: 0.9808 - val_loss: 0.8370 - val_acc: 0.8225
Epoch 00022: val_acc did not improve from 0.83292
Epoch 23/100
- 13s - loss: 0.0869 - acc: 0.9797 - val_loss: 0.7614 - val_acc: 0.8383
Epoch 00023: val_acc improved from 0.83292 to 0.83835, saving model to weights\weights_CNN_1.hdf5
Epoch 24/100
- 13s - loss: 0.0702 - acc: 0.9829 - val_loss: 0.9031 - val_acc: 0.8237
...
Epoch 00098: val_acc did not improve from 0.84879
Epoch 99/100
- 23s - loss: 1.3648 - acc: 0.9139 - val_loss: 3.1067 - val_acc: 0.8062
Epoch 00099: val_acc did not improve from 0.84879
Epoch 100/100
- 23s - loss: 1.4884 - acc: 0.9066 - val_loss: 3.0931 - val_acc: 0.8041
Epoch 00100: val_acc did not improve from 0.84879
CNN 2
Filter 6
Node 183
<keras.optimizers.Adagrad object at 0x7ff0d9147748>
Train on 9573 samples, validate on 2394 samples
Epoch 1/100
- 20s - loss: 15.6024 - acc: 0.0276 - val_loss: 15.6401 - val_acc: 0.0297
Epoch 00001: val_acc improved from -inf to 0.02966, saving model to weights\weights_CNN_2.hdf5
Epoch 2/100
- 13s - loss: 15.6736 - acc: 0.0276 - val_loss: 15.6401 - val_acc: 0.0297
Epoch 00002: val_acc did not improve from 0.02966
...
Epoch 99/100
- 13s - loss: 15.6736 - acc: 0.0276 - val_loss: 15.6401 - val_acc: 0.0297
Epoch 00099: val_acc did not improve from 0.02966
Epoch 100/100
- 13s - loss: 15.6736 - acc: 0.0276 - val_loss: 15.6401 - val_acc: 0.0297
Epoch 00100: val_acc did not improve from 0.02966
(2394, 9)
most_Accuracy of 9 models: [0.8187134502923976, 0.7443609022556391, 0.7059314954051796, 0.716374269005848, 0.8224728487886382, 0.8091060985797828, 0.8383458646616542, 0.8487886382623224, 0.029657477025898077]
most_Accuracy: 0.8872180451127819
most_F1_Micro: (0.8872180451127819, 0.8872180451127819, 0.8872180451127819, None)
most_F1_Macro: (0.883648009481651, 0.8806776773989405, 0.8813404998528928, None)
most_F1_weighted: (0.8893134179612591, 0.8872180451127819, 0.8874828407798586, None)
Process finished with exit code 0
I need to ask you two questions about the above results：
1.I now run 100 epochs per RDL. Does increasing the epoch number improve the accuracy of the model?
2.For example, what is the reason that the accuracy of CNN2 has not been improved?
Thanks ~
from rmdl.
zhuzhangli
commented on August 16, 2024
This is the plot of the 20NewsGroup dataset.
Looking forward to receiving your guidance~
from rmdl.
kk7nc
commented on August 16, 2024
Thank you for sending me the results,
The results which is sent is very similar to the paper we report,
Sometimes one or 2 models are not perform, but if we have more models the overall accuracy will be higher,
For this dataset, I think 100-150 epochs are enough for most of the RDLs
from rmdl.
zhuzhangli
commented on August 16, 2024
Ok, thank you~
from rmdl.
Related Issues (20)
- How to set the number of "n" in the experiment? HOT 3
- accuracy question
- accurcy problems HOT 11
- How to use saved model and weights for prediction? HOT 1
- Functional API HOT 1
- How to we make prediction on new data? HOT 1
- Error in model 0 try to re-generate an other model DNN 0 HOT 2
- Error in model 0 try to re-generate an other model DNN 0 HOT 4
- Upload Trained Model HOT 1
- Unseen data... HOT 1
- random_optimier not use
- Could not find %s Set GloVe Directory in Global.py HOT 1
- Multi label classifier in RMDL. HOT 1
- The dataset is not divided into train data, validation data and test data?
- Using my own image dataset HOT 1
- Error when trying to run for my dataset HOT 2
- How to use a trained model for predictions? HOT 3
- ValueError HOT 18
- Prediction time HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from rmdl.