This page answers two common questions about solution 2. Each answer is short. Each one links to the place in this chapter that explains it in full.
Contents#
- What does solution 2 add to solution 1, and what is a ranker?
- Why is solution 2 called the teacher, and why can a student not do better than its teacher?
1. What does solution 2 add to solution 1, and what is a ranker?#
Solution 2 lists every safe push for a glass, not just one. A ranker is a fitted model that puts such a list in order, best first. Here it is a set of small decision trees. It cannot add a push or bring back an unsafe one. See the model. On 50 tables it racked the same 107 glasses as the printed rule. It used 182 pushes, and the rule used 173.
2. Why is solution 2 called the teacher, and why can a student not do better than its teacher?#
Its recorded pushes are the examples that solutions 3, 6 and 7 copy. They cost only simulator time. A student that only copies sees only the pushes this solution can list. So it cannot learn a better push than these. See it is the teacher.
← How it compares · Imitation from demonstrations — what it is →