kk7nc / rmdl Goto Github PK

First of all thank you so much for your efforts in this paper, it is very interesting. But when I tried to run the code for text classification I got this error ("ValueError: max_features=[20, 500, 50], neither a positive integer nor None"). This error occurs when I run [RMDL.Text_Classification(X_train, y_train, X_test, y_test, batch_size, sparse_categorical, Random_Deep, n_epochs]. So can you please help me to overcome that.

Regards,
Saja

I found this library really helpful to my task of text classification, so thanks for turning it available. I would like to use other variables beside the text, though, as I can usually use in a Functional API model from Keras. Is it possible for me to make the changes necessary to apply this feature?

Why use test data as validation data directly?
Look forward to your reply. Thanks!

Hello, I want to classify images. How can I input for training my own dataset???

I am getting the following error while executing the following code.

from RMDL import RMDL_Text

text_classification=RMDL_Text.Text_Classification(Xtrain,Ytrain,Xtest,Ytest)

Could not find %s Set GloVe Directory in Global.py  <module 'RMDL.Download.Download_Glove' from 'C:\\Users\\Veni\\Anaconda3\\lib\\site-packages\\RMDL\\Download\\Download_Glove.py'>
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-22-a2b7412e3f8f> in <module>()
      1 from RMDL import RMDL_Text
      2 
----> 3 text_classification=RMDL_Text.Text_Classification(Xtrain,Ytrain,Xtest,Ytest)

~\Anaconda3\lib\site-packages\RMDL\RMDL_Text.py in Text_Classification(x_train, y_train, x_test, y_test, batch_size, EMBEDDING_DIM, MAX_SEQUENCE_LENGTH, MAX_NB_WORDS, GloVe_dir, GloVe_file, sparse_categorical, random_deep, epochs, plot, min_hidden_layer_dnn, max_hidden_layer_dnn, min_nodes_dnn, max_nodes_dnn, min_hidden_layer_rnn, max_hidden_layer_rnn, min_nodes_rnn, max_nodes_rnn, min_hidden_layer_cnn, max_hidden_layer_cnn, min_nodes_cnn, max_nodes_cnn, random_state, random_optimizor, dropout)
     56         if not os.path.isfile(GloVe_DIR):
     57             print("Could not find %s Set GloVe Directory in Global.py ", GloVe)
---> 58             exit()
     59 
     60     G.setup()

NameError: name 'exit' is not defined

I have an Image Data Set with dimensions 21x32x1. When I feed it into the model as described in the README.md i get this error. I have gone through all the previous threads related to this issue but it didn't solve my problem. My model never runs, it straight away throws this error.
Please guide me on how to solve it!

Hello, could you upload an already trained network such that one could use and test it without the time consuming training phase? This would be really great!

Hi.
I reproduced your code with PyTorch, but when I reviewed the code, I found that you didn't use random_optimizer when sorting images with CNN, but Adam instead.

Hi, I have trained model but I am not able to use trained weights and model for prediction( inference ) time.

Can you provide simple example for that?

Thank you !

I still met with this problem.
Error in model 0 try to re-generate an other model DNN 0
My dataset is very small.

Hello,

Your approach looks very interesting and I would like to test it on my own research. I've skimmed through the docs, readme and source code without finding an answer to this issue:

Once a model is trained, is there a way to save it as a file for future use? If so, how would you then reload it and use it to do predictions on new unlabelled data?

Describe the solution you'd like

Ideally, simple methods such as save_model() , load_model() and predict(), similar to what's already available on e.g. Keras or Scikit-Learn

Problem Statement
The problem I am solving is to classify genres of movies. So, the output in my case would be one-hot encoded for 20 genres(array of 20 elements with 1 or 0). For this we are using loss function of binary_crossentropy and sigmoid as the last layer activation.

Describe the solution you'd like
We would like to have an option of binary_crossentropy=True in parameters, and also number_of_classes should be configurable and not np.max(y_train)+1 which would always be 2 in my case.

Describe alternatives you've considered
No alternatives have been considered for now.

Hello, I am very interested in your article--An Improvement of Data Classification Using Random Multimodel Deep Learning (RMDL)
However, during the experiment, I found that the reproducibility accuracy of the four data sets WOS11967, WOS46985, IMDB and 20NewsGroup differed greatly. Is there any other information that needs attention besides the information published by github?
Looking forward to your guidance, thank you~

""Epoch 00001: val_acc improved from -inf to 0.76482, saving model to weights\weights_DNN_0.hdf5
Error in model 0 try to re-generate another model
DNN 0
<keras.optimizers.Adagrad object at 0x7f3339451dd8>
Train on 7769 samples, validate on 3019 samples
Epoch 1/120
Segmentation fault (core dumped)""
this appear when i start to run any examples.

This problem asked at kk7nc/HDLTex#5

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: [e.g. iOS]
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Smartphone (please complete the following information):

• Device: [e.g. iPhone6]
• OS: [e.g. iOS8.1]
• Browser [e.g. stock browser, safari]
• Version [e.g. 22]

Additional context
Add any other context about the problem here.

Hi

If I use RMDL and I want to predict unseen data, could I use predict function?
if yes, how?

Thanks

Hi, very interesting approach.
Can you provide some estimates of time performance for predict image classification.
For example how much time takes to predict 28x28 RGB image, it will be great to see comparison by VGG16
Thanks

Hello !
When I do some experiment on my own datasets according to your model,why the value of acc in some model is always 0.1000 from the begging to the end??
Thanks!

"""
batch_size = 100
sparse_categorical = 0
n_epochs = [200, 200, 300] ## DNN--RNN-CNN [500, 500, 500]
Random_Deep = [2, 2, 3] ## DNN--RNN-CNN
......
Epoch 00277: val_acc did not improve from 0.10000
Epoch 278/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00278: val_acc did not improve from 0.10000
Epoch 279/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00279: val_acc did not improve from 0.10000
Epoch 280/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00280: val_acc did not improve from 0.10000
Epoch 281/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00281: val_acc did not improve from 0.10000
Epoch 282/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00282: val_acc did not improve from 0.10000
Epoch 283/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00283: val_acc did not improve from 0.10000
Epoch 284/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00284: val_acc did not improve from 0.10000
Epoch 285/300

• 81s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00285: val_acc did not improve from 0.10000
Epoch 286/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00286: val_acc did not improve from 0.10000
Epoch 287/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00287: val_acc did not improve from 0.10000
Epoch 288/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00288: val_acc did not improve from 0.10000
Epoch 289/300

• 81s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00289: val_acc did not improve from 0.10000
Epoch 290/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00290: val_acc did not improve from 0.10000
Epoch 291/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00291: val_acc did not improve from 0.10000
Epoch 292/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00292: val_acc did not improve from 0.10000
Epoch 293/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00293: val_acc did not improve from 0.10000
Epoch 294/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00294: val_acc did not improve from 0.10000
Epoch 295/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00295: val_acc did not improve from 0.10000
Epoch 296/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00296: val_acc did not improve from 0.10000
Epoch 297/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00297: val_acc did not improve from 0.10000
Epoch 298/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00298: val_acc did not improve from 0.10000
Epoch 299/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00299: val_acc did not improve from 0.10000
Epoch 300/300

• 80s - loss: 14.5063 - acc: 0.1000 - val_loss: 14.5063 - val_acc: 0.1000

Epoch 00300: val_acc did not improve from 0.10000
(3000, 7)
(3000, 7)
Accuracy of 7 models: [0.485, 0.1, 0.718, 0.7346666666666667, 0.1, 0.6653333333333333, 0.1]
Accuracy: 0.6853333333333333
F1_Micro: 0.6853333333333333
F1_Macro: 0.6642739238169832
F1_weighted: 0.6642739238169834

Process finished with exit code 0
"""

Hello.I am very interested in this paper.I want to do some research in image classification in my own datasets,but i don't know how to set the n : Random_Deep = [3, 0, 3] ## DNN--RNN-CNN.
Does this value n set at random?

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.