The merits of multi-armed bandit testing against A/B testing are quickly becoming prominent in the AI research space with the growing availability of experimental data and the ability of researchers to quickly and easily run simultaneous and repeated trials and construct models against that data.
You might be familiar with A/B testing as a scientific process that tests a control population against a variant and compares the results of some process on the two groups, for example, an experimental medical drug that researchers are interested in measuring varying outcomes for.
Outside research sciences and fields such as healthcare, A/B testing is becoming increasingly popular with marketers in the online advertising space. In the next section we will fully explore how the process works.
Essentially, a multi-armed bandit approach to an...
