Summary
Thompson Sampling is a powerful sampling technique that enables you to quickly figure out the highest of a number of unknown conversion rates. It is always applied in the same frame, called the multi-armed bandit problem, which in the classic sense is composed of several slot machines, each one having a different conversion rate of positive outcomes. We had a first glance at how this AI solves this problem better and faster than standard methods.
In the next chapter, we will perform a full practical activity where we will see how the multi-armed bandit frame can easily model a business problem—online advertising—and how Thompson Sampling can bring significant added value.