enter image description here

I am trying to understand my loss curve using : tf.keras.losses.BinaryCrossentropy()

Question 1: Based on my loss curve/accuracy, would it be wise to proceed to feed it into a ensemble learning model alongside with other 10 similar binary classification model?

Question 2: I would know the reason why the apparent divergence between train / validation and whether is it a good/bad thing or any room for improvement/unrepresentative data

Question 3: (I understand mse/mae loss curve divergence/convergence; are they similar too in terms of interpretation?)

Tried looking around for learning loss curve but only found mse/mae interpretation so far but barely anything about BinaryCrossentropy. could someone help me please? TIA

  • 1
    $\begingroup$ It looks like your model is overfitting. What is the ratio between the positive class and the negative class? $\endgroup$
    – noe
    Mar 28 at 8:12
  • $\begingroup$ its 50/50 i ran a 20 patience val_loss callback & 800 epochs $\endgroup$ Mar 28 at 19:47
  • $\begingroup$ You may try to solve the overfitting. Maybe with dropout or regularization techniques. $\endgroup$
    – noe
    Mar 28 at 19:53


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.