Hands-On Deep Learning for Games

Testing through imitation

At this point in your learning, you have learned several strategies that we can apply to help our testing agent learn and find the goals. We can use curiosity or curriculum learning fairly easily, and we will leave that as an exercise for the reader. What we want is a way to control some of the testing process, and we don't really want our agent to randomly test everything (at least not at this stage). Sure, there are places where completely random testing works well. (By the way, this random form of testing is called monkey testing, because it resembles a monkey just mashing keys or input.) However, in a space such as our game, exploring every possible combination could take a very long time. Therefore, the best ...