Fine-tuning hyperparameters of a neural network
It is important to establish a baseline model before making any improvements in machine learning. A baseline model is a simple model that we can use to evaluate the performance of more complex models. In Chapter 5, Image Classification with Neural Networks, we achieved an accuracy of 88.50% on our training data and 85.67% on our test data in just five epochs. In our quest to try to improve our model’s performance, we will continue with our three-step (build, compile, and fit) process of constructing a neural network using TensorFlow. In each of the steps we use to build our neural network, there are settings that need to be configured before training. These settings are called hyperparameters. They control how the network will learn and perform, and mastering the art of fine-tuning them is an essential step in building successful deep learning models. Common hyperparameters include the number of neurons in each layer, the number...