Introduction

In the previous chapter, we continued to develop our knowledge of creating accurate models with neural networks by experimenting with cross-validation as a method to test how various hyperparameters perform in an unbiased manner. We utilized leave-one-out cross-validation, in which we leave one record out of the training process for use in validation and repeat this for every record in the dataset. Then, we looked at k-fold cross-validation, in which we split the training dataset into k folds, train the model on k-1 folds, and use the final fold for validation. These cross-validation methods allow us to train models with different hyperparameters and test their performance on unbiased data.

Deep learning is not only about building neural networks, training them using an available dataset, and reporting the model accuracy. It involves trying to understand your model and the dataset, as well as moving beyond a basic model by improving it in many aspects. In this chapter...