Callbacks in Keras are functions that can be run during the training process. They can do all kinds of great things, such as saving your model weights after an epoch, logging things, changing your hyperparameters, or conveniently writing TensorBoard log files. You can even create your own custom callbacks.

We will be using the TensorBoard callback in the next section; however, I'd encourage you to take a look at all the callbacks available in Keras at https://keras.io/callbacks.

TensorBoard callbacks are objects that can be configured and instantiated before model training.  We will be creating a list of these callbacks.  Once we have created a list of callbacks that we want to use with our deep neural network, we can simply pass that list as an argument to the model's .fit() method.   These callbacks will then be used at each epoch, or as appropriate by Keras.  This will make more sense as we walk through the next example.