CODEWITHSUNDEEP
pythondeep-learningkeras
Han
Han

Reputation: 311

Keras deep learning why validation accuracy stuck in a value every time?

I am trying to train InceptionV3 network with my custom dataset (36 classes 130 samples per each). And some parameters fro my network:

Upvotes: 0

Views: 4053

Answers (3)

Joonatan Samuel
Joonatan Samuel

Reputation: 651

Given just this information it is hard to tell what might be the underlying problem. In general, the machine learning engineer is always working with a direct trade-off between overfitting and model complexity. If the model isn't complex enough, it may not be powerful enough to capture all of the useful information necessary to solve a problem. However, if our model is very complex (especially if we have a limited amount of data at our disposal), we run the risk of overfitting. Deep learning takes the approach of solving very complex problems with complex models and taking additional countermeasures to prevent overfitting.

Three of the most common ways to do that are

  • Regularization
  • Dropout
  • Data augmentation

If your model is not complex enough:

  • Make it bigger (easy)
  • Make it smarter (hard)

Upvotes: 2

paolof89
paolof89

Reputation: 1369

Be more specific on the example, post the code you used to build the Sequential Model.

At the moment I can say that your problem could be in the initial dataset. You have 130 sample for 36 classes that means 3.6 example for each class?

Upvotes: 0

Andrey Lukyanenko
Andrey Lukyanenko

Reputation: 3851

It could mean that the model has learned everything possible and can't improve further.

One of the possible ways to improve accuracy is to get new data. You have ~4 samples per class, which is rather low. Try to get more samples or use data augmentation technics.

Upvotes: 1

Related Questions