Model selection criteria
webcsm opened this issue · comments
Hi @akshitac8 , congratulations on your work! Could you elaborate a bit on the model selection criteria? What losses do you pay attention to, and what behavior do you look for?
Hello, @webcsm thank you for your interest in our work. We selected the model using the harmonic mean criteria. We also observed the vae loss because if that loss is exploding then it means the model is not learning.
Thank you @akshitac8.
Just a quick follow up question. I'm assuming this harmonic mean criteria is computed on the test set, right?
It can be computed at validation set also during the training time. I have uploaded the code for computing it on the test data.