liveBook · Manning

Context

As explained in Applied Reinforcement Learning

Context is observable information about a decision situation that is available before an action is selected.

Context is observable information available before an action is chosen. In a contextual bandit, it helps select an action suited to the current situation rather than choosing one average best action for everyone. section 5.5

For a decision, context may include customer data, environmental conditions, or time-based signals. Unlike a changing state, context is observed before acting and does not change because of that action. section 5.5

Personalized discounts

At each customer interaction, use the available context to choose a discount level. The aim is to find offers that increase total revenue while balancing tests of new offers with use of proven ones. section 5.5

How Applied Reinforcement Learning covers it

How deep it goes
Defines context as a vector and distinguishes it from state, without detailing its construction.
Examples
Uses customer data, environmental conditions, time signals, and personalized discount levels as examples.
What you can do after
Apply observable customer information to choose discount levels for individual interactions.
How it connects
Connects context with contextual bandits, multi-armed bandits, reinforcement learning, and state.

Also explained in

Compare how 30 books explain context →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Applied Reinforcement Learning. The section links point to the parts of the book it draws on.