Inferring Models of Program Behavior Clause Samples

Inferring Models of Program Behavior. → ∈ ≡ ⇔ ∈ resp Fig. 1. Learning setting. function ▇▇▇▇ : Models Behaviors and we assume beh(P )= ▇▇▇▇ (model (P )). Within the area of model-based testing, this assumption is commonly referred to as the test hypothesis [11, 31]. Two models M and M′ are equivalent if they induce the same observable behavior: M M′ ▇▇▇▇ (M )= ▇▇▇▇ (M′). Many instantiations of this general framework are possible. In the case of reactive systems, for instance, the set Behaviors may consist of functions λ : Σ∗ Ω∗ from sequences of inputs to sequences of outputs that preserve the prefix ordering (here denoted ) and the length of sequences, that is, for all w, w′ Σ∗, w w′ λ(w) λ(w′) and λ(w) = w . In this case, the set Models naturally consists of (deterministic) ▇▇▇▇▇ machines with inputs Σ and outputs Ω. For reactive systems in which inputs and outputs do not alternate strictly, Behaviors may be defined as the class of prefix closed sets of suspension traces and Models as the class of I/O transition systems [69, 72].2 2.1 Model Learning