What is fascinating about deep learning is that we don't need to focus on feature engineering. Ideally, the network learns by itself what is important and what is not. An excellent example of such as case is behavioral cloning: a human demonstrates a certain task—this is the training data and an agent (deep learning model) tries to copy that behavior without specifying steps. In the case of autonomous vehicles in a protected environment with only a single agent, the agent should learn to drive on the road. We will demonstrate how to implement a deep learning model that teaches an agent to drive a car around a track.
Learning to drive a car with end-to-end learning
Getting started
For this recipe, we will be using...
