Improving Sample Efficiency with Normalized RBF Kernels

• data_load.py: Include functions related to data loading, pre-processing, and sampling.
• models.py: Include main clases for the models creation.
• params.py: Contain training configuration parameters and models hyper-parameters.
• train_model.py: Models training file with a single Center|Prototype by class.
• train_model_multiple_centers.py: Models training file with multiple Centers|Prototypes by class.
• silhouette_eval.py: Evaluation of the models Center|Prototypes distribution using the Silhouette coefficient.

For a simple command line execution of the training and the Silhouette evaluation files, reproduce as the following examples:

>> python  train_model.py  RESNET|CNN  CIFAR-10|CIFAR-100
>> python  train_model_multiple_centers.py  RESNET|CNN  CIFAR-10|CIFAR-100

>> python  silhouette_eval.py  RESNET|CNN  CIFAR-10|CIFAR-100