Compare CNN vs RNN classificators for sample terms
leandroradusky opened this issue · comments
Once created sample models for CNNs (#32) and RNNs (#36), select from the training set of the competition a sample term that contains enough positive and negative examples*, train them with both, and then compare their results as in #9, plotting ROC curves, comparing the AUC, etc.
This analysis should be done in a notebook in order to show the usage of the developed capabilities of the library.
- proteins that have the selected term assigned and others that don't, enough should be at least > 100 proteins, more will be better but they should fit in your RAM when training the set otherwise you will start to swap and things could go really slow.