In this presentation, James walkthroughs how Contextual Bandits and the Sequential Decision Making framework are helping him better understand his customers as a Senior Data Scientist at WooliesX.
Sequential Decision Making is an intuitive framework that we can use to map many problems that we face into an agent that acts and an environment which it interacts with. Learning to efficiently navigate these problems presents exciting opportunities in an active and exciting area of research: Decision Science.
The origins of these frameworks date back to temporal-difference learning papers Minky (1954) who may have been the first to release that these principles could be important for artificial learning systems. Exciting projects such as AlphaGo, AlphaStar, OpenAI Five and tremendous advancements in Atari Benchmarks through algorithmic innovation have brought tremendous activity to this area of research. Contextual Bandits are a particularly useful version of the Multi-Arm Bandit problem which aims to relate contextual information to reward distributions for various actions.
Want to sponsor? Contact us to find out more.