Building an RL Agent to complete tasks on the web – Call to Action

This recipe will teach you how to implement an RL training script so that you can train an RL Agent to handle Call-To-Action (CTA) type tasks for you. CTA buttons are the actionable buttons that you typically find on web pages that you need to click in order to proceed to the next step. While there are several CTA button examples available, some common examples include the OK/Cancel dialog boxes, where you need you to click to acknowledge/dismiss the pop-up notification, and the Click to learn more button. In this recipe, you will instantiate a RL training environment that provides visual rendering for the web pages containing a CTA task. You will be training a proximal policy optimization (PPO)-based deep RL Agent that's been implemented using TensorFlow 2.x to learn how to complete the task at hand.

The following image illustrates a set of observations from a randomized CTA environment (with different...