I'm training a deep learning model. After each epoch I measure the performance of the model on validation set. Here is how the performance looks like while training:

enter image description here

It's a binary classification task with cross entropy loss function. I use argmax at the last layer to do the prediction and measure precision and recall. Note, the number of positive and negative samples within each mini batch are almost the same (mini-batches are balanced). Any idea about possible reasons that the model is behaving like this? And how I can improve the recall as well as making it more stable like the precision?


  • $\begingroup$ I would start looking at the data first and evaluate the quality of the training samples and also the validation ones. I have seen this before, just making sure the quality and representativeness of the samples to the classes itself reduces the jumps and gets what you need. $\endgroup$ – Arun Aniyan Jul 24 '20 at 19:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.