Hands-On Q-Learning with Python

Other applications for bandit problems

One of the most well-known applications of the MABP methodology is in recommendation systems. For example, an advertiser's goal using a process such as A/B testing is to compare several alternative online advertisements and choose the one that performs best among users.

In A/B testing, we split our audience into similar groups and test different versions of ads on each group to compare their performance. In an advertising scenario, one online ad is treated as a control and another is treated as a test, similar to common experimental design in scientific testing. The control should be an ad of which the performance is known and can be measured against some established standard, and the test ads are then ...