Deep Learning with PyTorch

Evaluation protocol

Once you decide how you are going to evaluate the current progress, it is important to decide how you are going to evaluate on your dataset. We can choose from the three different ways of evaluating our progress:

• Holdout validation set: Most commonly used, particularly when you have enough data
• K-fold cross validation: When you have limited data, this strategy helps you to evaluate on different portions of the data, helping to give us a better view of the performance
• Iterated k-fold validation: When you are looking to go the extra mile with the performance of the model, this approach will help