Obtaining the best performing model using hyperparameter tuning

As described in Chapter 3, Developing a Powerful Deep Learning Model, obtaining a DL model that extracts the right pattern for the underlying task requires multiple components to be configured appropriately. While building the right model architecture often introduces many difficulties, setting up the proper model training is another challenge that most people struggle with.

In machine learning (ML), a hyperparameter refers to any parameter that controls the learning process. In many cases, data scientists often focus on model-relevant hyperparameters such as the number of a particular type of layer, learning rate, or type of optimizer. However, hyperparameters also include data-relevant configurations such as types of augmentation to apply and a sampling strategy for model training. The iterative process of changing a set of hyperparameters, and understanding performance changes, to find the right set of hyperparameters...