search

Serpeve / EEGSym

Open implementation and code from the publication "EEGSym: Overcoming Intersubject Variability in Motor Imagery Based BCIs with Deep Learning"
MIT License
21 stars 0 forks source link

Error creating model with pre-trained weights #3

Closed ShamimSasani closed 1 year ago

ShamimSasani commented 1 year ago

Hi, I have some problems during creating the model with the pretrained weights as follow : when I copy and paste the "Example of Use" without any change , firstly I got this error that the "model_hyparams" is not defined. then I defined it as model_hyparams= dict( ), and it continued. however, after reaching the "model.load_weights('EEGSym_pretrained_weights_{}_electrode.h5'.format(ncha))" I got this error: 

---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

[<ipython-input-5-5786c8fa12e0>](https://localhost:8080/#) in <module>
     50 # Load pre-trained weight values
     51 if pretrained:
---> 52     model.load_weights('EEGSym_pretrained_weights_{}_electrode.h5'.format(ncha))
     53 
     54 # Early stopping

2 frames

<__array_function__ internals> in transpose(*args, **kwargs)

[/usr/local/lib/python3.7/dist-packages/numpy/core/fromnumeric.py](https://localhost:8080/#) in _wrapfunc(obj, method, *args, **kwds)
     55 
     56     try:
---> 57         return bound(*args, **kwds)
     58     except TypeError:
     59         # A TypeError occurs if the object does have such a method in its

ValueError: axes don't match array

how can I fix this ?

another question: " why in the model summary, I got the dimension of the model's Input_layeras a 4 with "None" in the axis=0, is that correct? if it is correct, how can I make my x_train in this form ?

https://colab.research.google.com/drive/1b0LI54aT7p4sWQQ87EgDKEa0-8Hgts_9?usp=sharing

looking forward to hearing from you. shamim.

Serpeve commented 1 year ago

Hello Shamim,

Thank you for your interest in the open implementation of EEGSym. I am sorry for the error that you have encountered, I have updated the repository and now the code should not show you errors.

I have just updated it, solving the problem that you were getting and others that I found. The new code you can also creates a random signal and tests the whole process with pretrained weights for fine-tuning.

I have updated the requirements and tested the code for python==3.10.

I am sorry for the late response.

Can you confirm that there are no further issues?

Kind regards, Sergio Pérez-Velasco

ShamimSasani commented 1 year ago

Hi Serpeve, thank you for your response and also the update of the code. I will check them today and get back to you for confirmation about the fixed errors. can you give me a direct email so I can contact you for further questions as well? here is my email: shamim.sasani@gmail.com

best regards, shamim

ShamimSasani commented 1 year ago

I have checked the new example code presented in the README.md file available. and here is the results:

Model: "model"
__________________________________________________________________________________________________
 Layer (type)                   Output Shape         Param #     Connected to                     
==================================================================================================
 input_1 (InputLayer)           [(None, 384, 16, 1)  0           []                               
                                ]                                                                 

 tf.expand_dims (TFOpLambda)    (None, 1, 384, 16,   0           ['input_1[0][0]']                
                                1)                                                                

 tf.compat.v1.gather (TFOpLambd  (None, 1, 384, 7, 1  0          ['tf.expand_dims[0][0]']         
 a)                             )                                                                 

 tf.compat.v1.gather_2 (TFOpLam  (None, 1, 384, 2, 1  0          ['tf.expand_dims[0][0]']         
 bda)                           )                                                                 

 tf.compat.v1.gather_1 (TFOpLam  (None, 1, 384, 7, 1  0          ['tf.expand_dims[0][0]']         
 bda)                           )                                                                 

 concatenate (Concatenate)      (None, 1, 384, 9, 1  0           ['tf.compat.v1.gather[0][0]',    
                                )                                 'tf.compat.v1.gather_2[0][0]']  

 concatenate_1 (Concatenate)    (None, 1, 384, 9, 1  0           ['tf.compat.v1.gather_1[0][0]',  
                                )                                 'tf.compat.v1.gather_2[0][0]']  

 concatenate_2 (Concatenate)    (None, 2, 384, 9, 1  0           ['concatenate[0][0]',            
                                )                                 'concatenate_1[0][0]']          

 conv3d (Conv3D)                (None, 2, 384, 9, 2  408         ['concatenate_2[0][0]']          
                                4)                                                                

 conv3d_1 (Conv3D)              (None, 2, 384, 9, 2  792         ['concatenate_2[0][0]']          
                                4)                                                                

 conv3d_2 (Conv3D)              (None, 2, 384, 9, 2  1560        ['concatenate_2[0][0]']          
                                4)                                                                

 batch_normalization (BatchNorm  (None, 2, 384, 9, 2  96         ['conv3d[0][0]']                 
 alization)                     4)                                                                

 batch_normalization_1 (BatchNo  (None, 2, 384, 9, 2  96         ['conv3d_1[0][0]']               
 rmalization)                   4)                                                                

 batch_normalization_2 (BatchNo  (None, 2, 384, 9, 2  96         ['conv3d_2[0][0]']               
 rmalization)                   4)                                                                

 activation (Activation)        (None, 2, 384, 9, 2  0           ['batch_normalization[0][0]']    
                                4)                                                                

 activation_1 (Activation)      (None, 2, 384, 9, 2  0           ['batch_normalization_1[0][0]']  
                                4)                                                                

 activation_2 (Activation)      (None, 2, 384, 9, 2  0           ['batch_normalization_2[0][0]']  
                                4)                                                                

 dropout (Dropout)              (None, 2, 384, 9, 2  0           ['activation[0][0]']             
                                4)                                                                

 dropout_1 (Dropout)            (None, 2, 384, 9, 2  0           ['activation_1[0][0]']           
                                4)                                                                

 dropout_2 (Dropout)            (None, 2, 384, 9, 2  0           ['activation_2[0][0]']           
                                4)                                                                

 concatenate_3 (Concatenate)    (None, 2, 384, 9, 7  0           ['dropout[0][0]',                
                                2)                                'dropout_1[0][0]',              
                                                                  'dropout_2[0][0]']              

 add (Add)                      (None, 2, 384, 9, 7  0           ['concatenate_3[0][0]',          
                                2)                                'concatenate_2[0][0]']          

 average_pooling3d (AveragePool  (None, 2, 192, 9, 7  0          ['add[0][0]']                    
 ing3D)                         2)                                                                

 conv3d_3 (Conv3D)              (None, 2, 192, 1, 7  648         ['average_pooling3d[0][0]']      
                                2)                                                                

 batch_normalization_3 (BatchNo  (None, 2, 192, 1, 7  288        ['conv3d_3[0][0]']               
 rmalization)                   2)                                                                

 activation_3 (Activation)      (None, 2, 192, 1, 7  0           ['batch_normalization_3[0][0]']  
                                2)                                                                

 dropout_3 (Dropout)            (None, 2, 192, 1, 7  0           ['activation_3[0][0]']           
                                2)                                                                

 add_1 (Add)                    (None, 2, 192, 9, 7  0           ['average_pooling3d[0][0]',      
                                2)                                'dropout_3[0][0]']              

 conv3d_6 (Conv3D)              (None, 2, 192, 9, 2  6936        ['add_1[0][0]']                  
                                4)                                                                

 conv3d_7 (Conv3D)              (None, 2, 192, 9, 2  13848       ['add_1[0][0]']                  
                                4)                                                                

 conv3d_8 (Conv3D)              (None, 2, 192, 9, 2  27672       ['add_1[0][0]']                  
                                4)                                                                

 batch_normalization_6 (BatchNo  (None, 2, 192, 9, 2  96         ['conv3d_6[0][0]']               
 rmalization)                   4)                                                                

 batch_normalization_7 (BatchNo  (None, 2, 192, 9, 2  96         ['conv3d_7[0][0]']               
 rmalization)                   4)                                                                

 batch_normalization_8 (BatchNo  (None, 2, 192, 9, 2  96         ['conv3d_8[0][0]']               
 rmalization)                   4)                                                                

 activation_4 (Activation)      (None, 2, 192, 9, 2  0           ['batch_normalization_6[0][0]']  
                                4)                                                                

 activation_5 (Activation)      (None, 2, 192, 9, 2  0           ['batch_normalization_7[0][0]']  
                                4)                                                                

 activation_6 (Activation)      (None, 2, 192, 9, 2  0           ['batch_normalization_8[0][0]']  
                                4)                                                                

 dropout_4 (Dropout)            (None, 2, 192, 9, 2  0           ['activation_4[0][0]']           
                                4)                                                                

 dropout_5 (Dropout)            (None, 2, 192, 9, 2  0           ['activation_5[0][0]']           
                                4)                                                                

 dropout_6 (Dropout)            (None, 2, 192, 9, 2  0           ['activation_6[0][0]']           
                                4)                                                                

 concatenate_4 (Concatenate)    (None, 2, 192, 9, 7  0           ['dropout_4[0][0]',              
                                2)                                'dropout_5[0][0]',              
                                                                  'dropout_6[0][0]']              

 add_2 (Add)                    (None, 2, 192, 9, 7  0           ['concatenate_4[0][0]',          
                                2)                                'add_1[0][0]']                  

 average_pooling3d_1 (AveragePo  (None, 2, 96, 9, 72  0          ['add_2[0][0]']                  
 oling3D)                       )                                                                 

 conv3d_9 (Conv3D)              (None, 2, 96, 1, 72  648         ['average_pooling3d_1[0][0]']    
                                )                                                                 

 batch_normalization_9 (BatchNo  (None, 2, 96, 1, 72  288        ['conv3d_9[0][0]']               
 rmalization)                   )                                                                 

 activation_7 (Activation)      (None, 2, 96, 1, 72  0           ['batch_normalization_9[0][0]']  
                                )                                                                 

 dropout_7 (Dropout)            (None, 2, 96, 1, 72  0           ['activation_7[0][0]']           
                                )                                                                 

 add_3 (Add)                    (None, 2, 96, 9, 72  0           ['average_pooling3d_1[0][0]',    
                                )                                 'dropout_7[0][0]']              

 conv3d_12 (Conv3D)             (None, 2, 96, 9, 36  41508       ['add_3[0][0]']                  
                                )                                                                 

 conv3d_15 (Conv3D)             (None, 2, 96, 9, 36  2592        ['add_3[0][0]']                  
                                )                                                                 

 batch_normalization_12 (BatchN  (None, 2, 96, 9, 36  144        ['conv3d_12[0][0]']              
 ormalization)                  )                                                                 

 batch_normalization_15 (BatchN  (None, 2, 96, 9, 36  144        ['conv3d_15[0][0]']              
 ormalization)                  )                                                                 

 activation_8 (Activation)      (None, 2, 96, 9, 36  0           ['batch_normalization_12[0][0]'] 
                                )                                                                 

 activation_9 (Activation)      (None, 2, 96, 9, 36  0           ['batch_normalization_15[0][0]'] 
                                )                                                                 

 dropout_8 (Dropout)            (None, 2, 96, 9, 36  0           ['activation_8[0][0]']           
                                )                                                                 

 dropout_9 (Dropout)            (None, 2, 96, 9, 36  0           ['activation_9[0][0]']           
                                )                                                                 

 add_4 (Add)                    (None, 2, 96, 9, 36  0           ['dropout_8[0][0]',              
                                )                                 'dropout_9[0][0]']              

 average_pooling3d_2 (AveragePo  (None, 2, 48, 9, 36  0          ['add_4[0][0]']                  
 oling3D)                       )                                                                 

 conv3d_14 (Conv3D)             (None, 2, 48, 1, 36  11664       ['average_pooling3d_2[0][0]']    
                                )                                                                 

 batch_normalization_14 (BatchN  (None, 2, 48, 1, 36  144        ['conv3d_14[0][0]']              
 ormalization)                  )                                                                 

 activation_10 (Activation)     (None, 2, 48, 1, 36  0           ['batch_normalization_14[0][0]'] 
                                )                                                                 

 dropout_10 (Dropout)           (None, 2, 48, 1, 36  0           ['activation_10[0][0]']          
                                )                                                                 

 add_5 (Add)                    (None, 2, 48, 9, 36  0           ['average_pooling3d_2[0][0]',    
                                )                                 'dropout_10[0][0]']             

 conv3d_16 (Conv3D)             (None, 2, 48, 9, 36  10404       ['add_5[0][0]']                  
                                )                                                                 

 conv3d_19 (Conv3D)             (None, 2, 48, 9, 36  1296        ['add_5[0][0]']                  
                                )                                                                 

 batch_normalization_16 (BatchN  (None, 2, 48, 9, 36  144        ['conv3d_16[0][0]']              
 ormalization)                  )                                                                 

 batch_normalization_19 (BatchN  (None, 2, 48, 9, 36  144        ['conv3d_19[0][0]']              
 ormalization)                  )                                                                 

 activation_11 (Activation)     (None, 2, 48, 9, 36  0           ['batch_normalization_16[0][0]'] 
                                )                                                                 

 activation_12 (Activation)     (None, 2, 48, 9, 36  0           ['batch_normalization_19[0][0]'] 
                                )                                                                 

 dropout_11 (Dropout)           (None, 2, 48, 9, 36  0           ['activation_11[0][0]']          
                                )                                                                 

 dropout_12 (Dropout)           (None, 2, 48, 9, 36  0           ['activation_12[0][0]']          
                                )                                                                 

 add_6 (Add)                    (None, 2, 48, 9, 36  0           ['dropout_11[0][0]',             
                                )                                 'dropout_12[0][0]']             

 average_pooling3d_3 (AveragePo  (None, 2, 24, 9, 36  0          ['add_6[0][0]']                  
 oling3D)                       )                                                                 

 conv3d_18 (Conv3D)             (None, 2, 24, 1, 36  11664       ['average_pooling3d_3[0][0]']    
                                )                                                                 

 batch_normalization_18 (BatchN  (None, 2, 24, 1, 36  144        ['conv3d_18[0][0]']              
 ormalization)                  )                                                                 

 activation_13 (Activation)     (None, 2, 24, 1, 36  0           ['batch_normalization_18[0][0]'] 
                                )                                                                 

 dropout_13 (Dropout)           (None, 2, 24, 1, 36  0           ['activation_13[0][0]']          
                                )                                                                 

 add_7 (Add)                    (None, 2, 24, 9, 36  0           ['average_pooling3d_3[0][0]',    
                                )                                 'dropout_13[0][0]']             

 conv3d_20 (Conv3D)             (None, 2, 24, 9, 18  2610        ['add_7[0][0]']                  
                                )                                                                 

 conv3d_23 (Conv3D)             (None, 2, 24, 9, 18  648         ['add_7[0][0]']                  
                                )                                                                 

 batch_normalization_20 (BatchN  (None, 2, 24, 9, 18  72         ['conv3d_20[0][0]']              
 ormalization)                  )                                                                 

 batch_normalization_23 (BatchN  (None, 2, 24, 9, 18  72         ['conv3d_23[0][0]']              
 ormalization)                  )                                                                 

 activation_14 (Activation)     (None, 2, 24, 9, 18  0           ['batch_normalization_20[0][0]'] 
                                )                                                                 

 activation_15 (Activation)     (None, 2, 24, 9, 18  0           ['batch_normalization_23[0][0]'] 
                                )                                                                 

 dropout_14 (Dropout)           (None, 2, 24, 9, 18  0           ['activation_14[0][0]']          
                                )                                                                 

 dropout_15 (Dropout)           (None, 2, 24, 9, 18  0           ['activation_15[0][0]']          
                                )                                                                 

```add_8 (Add)                    (None, 2, 24, 9, 18  0           ['dropout_14[0][0]',             
                                )                                 'dropout_15[0][0]']             

 average_pooling3d_4 (AveragePo  (None, 2, 12, 9, 18  0          ['add_8[0][0]']                  
 oling3D)                       )                                                                 

 conv3d_22 (Conv3D)             (None, 2, 12, 1, 18  2916        ['average_pooling3d_4[0][0]']    
                                )                                                                 

 batch_normalization_22 (BatchN  (None, 2, 12, 1, 18  72         ['conv3d_22[0][0]']              
 ormalization)                  )                                                                 

 activation_16 (Activation)     (None, 2, 12, 1, 18  0           ['batch_normalization_22[0][0]'] 
                                )                                                                 

 dropout_16 (Dropout)           (None, 2, 12, 1, 18  0           ['activation_16[0][0]']          
                                )                                                                 

 add_9 (Add)                    (None, 2, 12, 9, 18  0           ['average_pooling3d_4[0][0]',    
                                )                                 'dropout_16[0][0]']             

 conv3d_24 (Conv3D)             (None, 2, 12, 9, 18  1296        ['add_9[0][0]']                  
                                )                                                                 

 batch_normalization_24 (BatchN  (None, 2, 12, 9, 18  72         ['conv3d_24[0][0]']              
 ormalization)                  )                                                                 

 activation_17 (Activation)     (None, 2, 12, 9, 18  0           ['batch_normalization_24[0][0]'] 
                                )                                                                 

 dropout_17 (Dropout)           (None, 2, 12, 9, 18  0           ['activation_17[0][0]']          
                                )                                                                 

 add_10 (Add)                   (None, 2, 12, 9, 18  0           ['add_9[0][0]',                  
                                )                                 'dropout_17[0][0]']             

 average_pooling3d_5 (AveragePo  (None, 2, 6, 9, 18)  0          ['add_10[0][0]']                 
 oling3D)                                                                                         

 conv3d_25 (Conv3D)             (None, 1, 6, 1, 18)  5832        ['average_pooling3d_5[0][0]']    

 batch_normalization_25 (BatchN  (None, 1, 6, 1, 18)  72         ['conv3d_25[0][0]']              
 ormalization)                                                                                    

 activation_18 (Activation)     (None, 1, 6, 1, 18)  0           ['batch_normalization_25[0][0]'] 

 dropout_18 (Dropout)           (None, 1, 6, 1, 18)  0           ['activation_18[0][0]']          

 add_11 (Add)                   (None, 2, 6, 9, 18)  0           ['average_pooling3d_5[0][0]',    
                                                                  'dropout_18[0][0]']             

 conv3d_26 (Conv3D)             (None, 1, 6, 1, 18)  5832        ['add_11[0][0]']                 

 batch_normalization_26 (BatchN  (None, 1, 6, 1, 18)  72         ['conv3d_26[0][0]']              
 ormalization)                                                                                    

 activation_19 (Activation)     (None, 1, 6, 1, 18)  0           ['batch_normalization_26[0][0]'] 

 dropout_19 (Dropout)           (None, 1, 6, 1, 18)  0           ['activation_19[0][0]']          

 add_12 (Add)                   (None, 2, 6, 9, 18)  0           ['add_11[0][0]',                 
                                                                  'dropout_19[0][0]']             

 conv3d_27 (Conv3D)             (None, 1, 6, 1, 18)  648         ['add_12[0][0]']                 

 batch_normalization_27 (BatchN  (None, 1, 6, 1, 18)  72         ['conv3d_27[0][0]']              
 ormalization)                                                                                    

 activation_20 (Activation)     (None, 1, 6, 1, 18)  0           ['batch_normalization_27[0][0]'] 

 dropout_20 (Dropout)           (None, 1, 6, 1, 18)  0           ['activation_20[0][0]']          

 conv3d_28 (Conv3D)             (None, 1, 1, 1, 18)  1944        ['dropout_20[0][0]']             

 batch_normalization_28 (BatchN  (None, 1, 1, 1, 18)  72         ['conv3d_28[0][0]']              
 ormalization)                                                                                    

 activation_21 (Activation)     (None, 1, 1, 1, 18)  0           ['batch_normalization_28[0][0]'] 

 dropout_21 (Dropout)           (None, 1, 1, 1, 18)  0           ['activation_21[0][0]']          

 add_13 (Add)                   (None, 1, 6, 1, 18)  0           ['dropout_20[0][0]',             
                                                                  'dropout_21[0][0]']             

 conv3d_29 (Conv3D)             (None, 1, 1, 1, 36)  216         ['add_13[0][0]']                 

 batch_normalization_29 (BatchN  (None, 1, 1, 1, 36)  144        ['conv3d_29[0][0]']              
 ormalization)                                                                                    

 activation_22 (Activation)     (None, 1, 1, 1, 36)  0           ['batch_normalization_29[0][0]'] 

 dropout_22 (Dropout)           (None, 1, 1, 1, 36)  0           ['activation_22[0][0]']          

 conv3d_30 (Conv3D)             (None, 1, 1, 1, 36)  1296        ['dropout_22[0][0]']             

 batch_normalization_30 (BatchN  (None, 1, 1, 1, 36)  144        ['conv3d_30[0][0]']              
 ormalization)                                                                                    

 activation_23 (Activation)     (None, 1, 1, 1, 36)  0           ['batch_normalization_30[0][0]'] 

 dropout_23 (Dropout)           (None, 1, 1, 1, 36)  0           ['activation_23[0][0]']          

 add_14 (Add)                   (None, 1, 1, 1, 36)  0           ['dropout_22[0][0]',             
                                                                  'dropout_23[0][0]']             

 conv3d_31 (Conv3D)             (None, 1, 1, 1, 36)  1296        ['add_14[0][0]']                 

 batch_normalization_31 (BatchN  (None, 1, 1, 1, 36)  144        ['conv3d_31[0][0]']              
 ormalization)                                                                                    

 activation_24 (Activation)     (None, 1, 1, 1, 36)  0           ['batch_normalization_31[0][0]'] 

 dropout_24 (Dropout)           (None, 1, 1, 1, 36)  0           ['activation_24[0][0]']          

 add_15 (Add)                   (None, 1, 1, 1, 36)  0           ['add_14[0][0]',                 
                                                                  'dropout_24[0][0]']             

 conv3d_32 (Conv3D)             (None, 1, 1, 1, 36)  1296        ['add_15[0][0]']                 

 batch_normalization_32 (BatchN  (None, 1, 1, 1, 36)  144        ['conv3d_32[0][0]']              
 ormalization)                                                                                    

 activation_25 (Activation)     (None, 1, 1, 1, 36)  0           ['batch_normalization_32[0][0]'] 

 dropout_25 (Dropout)           (None, 1, 1, 1, 36)  0           ['activation_25[0][0]']          

 add_16 (Add)                   (None, 1, 1, 1, 36)  0           ['add_15[0][0]',                 
                                                                  'dropout_25[0][0]']             

 conv3d_33 (Conv3D)             (None, 1, 1, 1, 36)  1296        ['add_16[0][0]']                 

 batch_normalization_33 (BatchN  (None, 1, 1, 1, 36)  144        ['conv3d_33[0][0]']              
 ormalization)                                                                                    

 activation_26 (Activation)     (None, 1, 1, 1, 36)  0           ['batch_normalization_33[0][0]'] 

 dropout_26 (Dropout)           (None, 1, 1, 1, 36)  0           ['activation_26[0][0]']          

 add_17 (Add)                   (None, 1, 1, 1, 36)  0           ['add_16[0][0]',                 
                                                                  'dropout_26[0][0]']             

 flatten (Flatten)              (None, 36)           0           ['add_17[0][0]']                 

 dense (Dense)                  (None, 2)            74          ['flatten[0][0]']                

==================================================================================================
Total params: 162,152
Trainable params: 160,496
Non-trainable params: 1,656
__________________________________________________________________________________________________
Epoch 1/500
18/18 [==============================] - 204s 11s/step - loss: 1.6567 - accuracy: 0.5083 - val_loss: 0.7143 - val_accuracy: 0.4550
Epoch 2/500
18/18 [==============================] - 190s 10s/step - loss: 1.6622 - accuracy: 0.4833 - val_loss: 0.7552 - val_accuracy: 0.5000
Epoch 3/500
18/18 [==============================] - 189s 10s/step - loss: 1.6357 - accuracy: 0.4783 - val_loss: 0.8154 - val_accuracy: 0.5100
Epoch 4/500
18/18 [==============================] - 187s 10s/step - loss: 1.5270 - accuracy: 0.5050 - val_loss: 0.8799 - val_accuracy: 0.5250
Epoch 5/500
18/18 [==============================] - 185s 10s/step - loss: 1.4210 - accuracy: 0.5000 - val_loss: 0.9373 - val_accuracy: 0.5200
Epoch 6/500
18/18 [==============================] - 183s 10s/step - loss: 1.4004 - accuracy: 0.4967 - val_loss: 0.9849 - val_accuracy: 0.5250
Epoch 7/500
18/18 [==============================] - 181s 10s/step - loss: 1.3900 - accuracy: 0.4883 - val_loss: 1.0248 - val_accuracy: 0.5250
Epoch 8/500
18/18 [==============================] - 177s 9s/step - loss: 1.3331 - accuracy: 0.4850 - val_loss: 1.0522 - val_accuracy: 0.5200
Epoch 9/500
18/18 [==============================] - 176s 9s/step - loss: 1.3708 - accuracy: 0.4733 - val_loss: 1.0724 - val_accuracy: 0.5150
Epoch 10/500
18/18 [==============================] - 176s 9s/step - loss: 1.2359 - accuracy: 0.4883 - val_loss: 1.0799 - val_accuracy: 0.5050
Epoch 11/500
18/18 [==============================] - 175s 9s/step - loss: 1.2221 - accuracy: 0.5000 - val_loss: 1.0771 - val_accuracy: 0.5050
Epoch 12/500
18/18 [==============================] - 174s 9s/step - loss: 1.0934 - accuracy: 0.5233 - val_loss: 1.0807 - val_accuracy: 0.5000
Epoch 13/500
18/18 [==============================] - 175s 9s/step - loss: 1.1810 - accuracy: 0.4900 - val_loss: 1.0726 - val_accuracy: 0.5000
Epoch 14/500
18/18 [==============================] - 175s 9s/step - loss: 1.1158 - accuracy: 0.5117 - val_loss: 1.0626 - val_accuracy: 0.5000
Epoch 15/500
18/18 [==============================] - 177s 9s/step - loss: 1.0876 - accuracy: 0.4967 - val_loss: 1.0460 - val_accuracy: 0.5000
Epoch 16/500
18/18 [==============================] - 176s 9s/step - loss: 0.9918 - accuracy: 0.5300 - val_loss: 1.0278 - val_accuracy: 0.5000
Epoch 17/500
18/18 [==============================] - 183s 10s/step - loss: 0.9795 - accuracy: 0.5250 - val_loss: 1.0102 - val_accuracy: 0.4950
Epoch 18/500
18/18 [==============================] - 186s 10s/step - loss: 0.9753 - accuracy: 0.5283 - val_loss: 0.9909 - val_accuracy: 0.4900
Epoch 19/500
18/18 [==============================] - 188s 10s/step - loss: 0.9106 - accuracy: 0.5283 - val_loss: 0.9738 - val_accuracy: 0.5050
Epoch 20/500
18/18 [==============================] - 193s 10s/step - loss: 0.9821 - accuracy: 0.5217 - val_loss: 0.9567 - val_accuracy: 0.4900
Epoch 21/500
18/18 [==============================] - 182s 10s/step - loss: 0.9361 - accuracy: 0.4883 - val_loss: 0.9362 - val_accuracy: 0.4900
Epoch 22/500
18/18 [==============================] - 183s 10s/step - loss: 0.9371 - accuracy: 0.5017 - val_loss: 0.9185 - val_accuracy: 0.4800
Epoch 23/500
18/18 [==============================] - 179s 10s/step - loss: 0.8806 - accuracy: 0.5133 - val_loss: 0.9018 - val_accuracy: 0.4800
Epoch 24/500
18/18 [==============================] - 177s 9s/step - loss: 0.8786 - accuracy: 0.4967 - val_loss: 0.8869 - val_accuracy: 0.4800
Epoch 25/500
18/18 [==============================] - 175s 9s/step - loss: 0.8668 - accuracy: 0.5033 - val_loss: 0.8689 - val_accuracy: 0.4850
Epoch 26/500
19/18 [==============================] - ETA: -2s - loss: 0.8193 - accuracy: 0.5233Restoring model weights from the end of the best epoch: 1.
18/18 [==============================] - 175s 9s/step - loss: 0.8193 - accuracy: 0.5233 - val_loss: 0.8558 - val_accuracy: 0.4900
Epoch 26: early stopping
7/7 [==============================] - 11s 2s/step

so far it got no errors and this is the best it got and then eary_stoppingworked.

I am training the model with a real small data set just now, I've removed the callbacks=[early_stopping] and reduced the epochsto 200 for faster results. dose it going to lower the expected accuracy you mentioned in the paper? because the other hyperparameters are the same as you set them.

looking forward to hearing from you. Best regards, shamim.