How to study the effect of each information on a deep neural network model?

Question

I am working on training a neural network model using the Python and Keras libraries.

My model testing accuracy is very low (60.0%), and I tried to raise it a lot, but I could not. I use the DEAP dataset (total 32 members) to train the model. The separation technique that I use is fixed. It was as follows: 28 participants for training, 2 for testing and 2 for testing.

For the model I am using is as follows.

• sequential model
• Optimizer = Adam
• With L2_regulator, Gaussian noise, dropout and batch normalization
• The number of hidden layers = 3
• Activation = relu
• Compilation error = categorical_processor
• initializer = he_normal

( ), , . , . , , () () ?

,

Answer 1

: CIFAR-10, :

• (-, ),
• ( ).

, , ( , ).

, ( Neptune, Jupyter Notebook TensorBoard ):

, , :

: deepsense.ai, - .