Suppose we have a scenario where we need to choose from a set of actions
Let's say we have 2 news articles:
We will show how to use the decision service to decide on the best article based on the user's location
1. Make a Decision
The decision service makes it easy to make smarter decisions using the context you have available.
Let's apply the decision service to a smart news application. The first step is to use the context available to make a decision.
The decision service returns an "action". In our example, this is a news article that was chosen based on the location.
2. Assign a Reward
We now need to report whether the action was successful or not so the decision service can learn to make better decisions. If this user clicks on the article, a reward of 1 is sent. Otherwise, a reward of 0 is sent.
3. Repeat and make smarter decisions
The decision service uses the information above to learn a model that maximizes reward. By collecting more and more data, the service continuously improves upon the existing model.