The following are the questions:
- How is TensorFlow Lite different from usual TensorFlow?
- Can you try and if you can build the model on a movie review dataset in  Chapter 3, Sentiment Analysis in Your Browser Using TensorFlow.js? Do you face some issues with Tensorflow Lite in that case?
- Can you try Adam optimizer and see if it improves the performance of the model?
- Can you think you operations other than Dropout which are also not important for inference on mobile?