Summary of IMDB

Now you should have a better idea of how to go about processing natural language texts and dialogues through a simple feedforward neural network. In this subsection of our journey, you learned how to execute a binary sentiment classification task using a feedforward neural network. In doing so, you learned how to pad and vectorize your natural language data, preparing it for processing with neural networks. You also went over the key architectural changes that are involved in binary classification, such as using an output neuron and the sigmoid activation function on the last layer of our network. You also saw how you can leverage a validation split in your data to get an idea of how your model performs on unseen data after each training epoch. Moreover, you learned how to indirectly interact with your model during the training process by using Keras callbacks. Callbacks can be useful for a variety of use cases, ranging from saving your model at a certain checkpoint or terminating the training session when a desired metric has reached a certain point. We can use the history callback to visualize training statistics, and we can use the early stopping callback to designate a moment to terminate the current training session. Finally, you saw how you can probe your network's predictions per review to better understand what kind of mistakes it makes:

• Exercise: Improve performance with regularization, as we did in the MNIST example.