Part 2 Look ma, no hands! Deep networks in the real world

A well-versed ML practitioner is a multifaceted individual. Not only do they need a good understanding of modern deep learning frameworks such as TensorFlow, but they also need to be able to navigate the complex APIs offered by it to implement complex deep learning models to solve some of the common machine learning problems in domains like computer vision and natural language processing.

In part 2, we look at real-world problems in both computer vision and natural language processing. First, we look at image classification and image segmentation, which are two popular computer vision tasks. For these tasks, we analyze modern complex deep learning models that have performed well on a given problem. Not only will we implement these models from scratch, we will understand the reasoning behind core design decisions and the advantages they bring about.

Next, we move to natural language processing. We first look at a sentiment analysis task and how deep learning can solve it. We also explore various corners of the solution, such as basic NLP preprocessing steps and using word vectors to enhance performance. We then look at language modeling: a pretraining task that gives the enormous language understanding enjoyed by state-of-the-art NLP models. In this discussion, we again discuss various techniques that are incorporated in language modeling to enhance the prediction quality.