Promise problem
Appearance
In computability theory, a promise problem is a generalization of a decision problem. It is defined by two decision problems L1 and L2 with L1 ∩ L2 = ∅. A Turing machine decides a promise problem if, for any x ∈ L1 ∪ L2, it accepts x if x ∈ L1 and rejects x if x ∈ L2. There are no requirements on the behavior of the Turing machine when x ∉ L1 ∪ L2 (this is the promise: that x is in one of the two sets).
If L2 is equal to Γ+ \ L1, the set of all words not in L1, then this is just the decision problem for L1.