13.4.1 Experimental Settings

We evaluated our proposal on an Intel Xeon E5 v3 computer with 128 GB of RAM using an NVIDIA Titan Xp GPU for both training and testing. All of the CNN architectures were implemented using Python3 under the Keras library and Tensorflow as the backend.

To measure the performance of both our approaches, we used the F1-score and the accuracy metrics, as shown in the equations below. We chose th...