CS345 PROBLEM SET #2
SOLUTIONS
6 Questions, 100 pts.

Problem 1 (25 pts.)

In this problem we will use two kinds of electronic gates, NAND gates and INVOR gates. A NAND gate has inputs X and Y; its output Z is F(alse) if and only if both X and Y are T(rue). In this problem the only INVOR gate has inputs X and Y; its output Z is F(alse) if and only if X is T(rue) and Y is F(alse). We can represents a circuit of NAND and INVOR gates by the EDB predicate wireT(Z), meaning that wire Z is a circuit input set to T(rue), nandGate(X,Y,Z) meaning that there is a NAND gate in the circuit with input wires X and Y and with output wire Z, and invorGate(X,Y,Z) meaning that there is an INVOR gate in the circuit with input wires X and Y and with output wire Z. The rules describing a circuit are then the following.
t(Z) :- wireT(Z).
t(Z) :- nandGate(X,Y,Z) & NOT t(X).
t(Z) :- nandGate(X,Y,Z) & NOT t(Y).
t(Z) :- invorGate(X,Y,Z) & NOT t(X).
t(Z) :- invorGate(X,Y,Z) & t(Y).
Here is a circuit of NAND and INVOR gates represented by the EDB: wireT(1), nandGate(6,1,3), nandGate(1,2,4), nandGate(3,5,6), nandGate(1,4,7), and invorGate(3,1,5).

For this EDB and the logic program above it, find the well-founded model by repeatedly making positive inferences and looking for unfounded sets.

Solution

Using the first rule we infer t(1) from t(1) :- wireT(1). Using the last rule we infer t(5) from t(5) :- invorGate(3,1,5) & t(1). Then, we instantiate the rules in all possible ways and eliminate those with a known false subgoal. For convenience, we also eliminate true subgoals from the remaining bodies and rules whose head has already been inferred. The result is
t(3) :- NOT t(6).
t(4) :- NOT t(2).
t(6) :- NOT t(3).
t(7) :- NOT t(4).
The largest unfounded set is {t(2)} so we infer NOT t(2). Then we infer t(4) from t(4) :- NOT t(2). After eliminating the last rule because NOT t(4) is F(alse) the remaining rules are
t(3) :- NOT t(6).
t(6) :- NOT t(3).
The largest unfounded set is {t(7)} so we infer NOT t(7). The remaining rules are the same and there are no unfounded sets so we are done. The WF model is {t(1),NOT t(2),t(4),t(5),NOT t(7)}. The value of t(3) and t(6) is "unknown".

Problem 2 (15pts.)

For the same EDB and program as in Problem 1, find the well-founded model by the alternating fixedpoint method.

Solution

The table below illustrates finding the well-founded model by the alternating fixedpoint method.

Round 0 1 2 3 4
t(1) F T T T T
t(2) F F F F F
t(3) F T F T F
t(4) F T T T T
t(5) F T T T T
t(6) F T F T F
t(7) F T F F F

Note that column 4 is the same as column 2 so we don't need to do any more iterations. The WF model is {t(1),NOT t(2),t(4),t(5),NOT t(7)} because t(1),t(4), and t(5) converge to T(rue), t(2) and t(7) converge to F(alse), and t(3) and t(6) oscillate.

Problem 3 (20pts.)

For the EDB and program of Problem 1, find all the stable models. Is there a stable semantics for this EDB and program?

Solution

There are only two stable models: {t(1),NOT t(2),NOT t(3),t(4),t(5),t(6),NOT t(7)} and {t(1),NOT t(2),t(3),t(4),t(5),NOT t(6),NOT t(7)}. There is no stable semantics for the given EDB and program because there is no unique minimal model.

Problem 4 (15pts.)

Suppose we change the input of terminal 1 to be F(alse); i.e., the relation wireT is made empty instead of {1}. Show that this EDB plus the program from Problem 1 is modularly stratified. What are the "modules"?

Solution

To be supplied later

Problem 5 (10pts.)

Is the program and EDB of Problem 4 locally stratified? Justify your answer.

Solution

The program and EDB of Problem 4 is not locally stratified because there is a negative cycle between the ground atoms t(3) and t(6). The cycle arises from the following two rule instantiations.
t(3) :- nandGate(6,1,3) & NOT t(6).
t(6) :- nandGate(3,5,6) & NOT t(3).

Problem 6 (15pts.)

Compare the well-founded and stable approaches on the following rules of propositional calculus.
p :- NOT q.
q :- NOT r.
r :- NOT s.
s :- NOT p.

Solution

The WF model for the above rules is {}, i.e., the value of p,q,r, and s is "unknown". There is no stable semantics for the above rules because there are two minimal stable models {p,NOT q,r,NOT s} and {NOT p,q,NOT r,s}.