Could you please help me understand why validation_loss increases to and ends with 1.823 on the final epoch? The validation accuracy seems to increase normally.
Hi @JeainnyKim ,
loss and accuracy are not always inversely correlated. Loss depends on the difference between raw prediction and class, while accuracy measures the difference between the prediction converted to 0 or 1 (the thresholded prediction) and the class. Accuracy and loss are still inversely correlated most of the times, but it can happen that they are not, like probably in your case.