We study the following gray-box learning problem: Given the serial composition of two Mealy machines A and B, where A is known and B is unknown, the goal is to learn a model of B using only output and equivalence queries on the composed machine. We introduce an algorithm that solves this problem, using at most |B| equivalence queries, independently of the size of A. We discuss its efficient implementation and evaluate the algorithm on existing benchmark sets as well as randomly-generated machines.
展开▼
机译：我们研究以下灰盒学习问题：给定两个Mealy机器A和B的序列组成，其中A已知，而B未知，目标是仅使用组成机器上的输出和等效查询来学习B的模型。我们介绍一种最多可使用| B |来解决此问题的算法。等价查询，与A的大小无关。我们讨论了它的有效实现，并在现有基准集以及随机生成的计算机上评估了该算法。
展开▼
