Problem set graph isomorphism question—use a reduction?
Probably not, this problem set doesn’t need you to do reductions
NP Completeness from Circuit Satisfiability
Summary of section 34.4
Formula satisfiability instead of circuit satisfiability
Boolean variables connected with logic operations
Formula satisfiability proved NP complete
CIRCUIT-SAT ≤p SAT means what about relative complexities?
Really only that CIRCUIT-SAT is no more than polynomially more complex than SAT
Particular reduction tells you something about the polynomial in
“polynomially more,” but not that there isn’t a
smaller polynomial in a cleverer reduction
3-CNF-SAT
Conjunctive normal form = formula is conjunction of disjunctions
3-literal clauses combined by and
Also NP complete
NP-hard vs NP-complete?
NP-complete = everything in NP reduces to it, and in NP
NP-hard = everything in NP reduces to it, not necessarily in NP itself
Are there things not in NP and if so what?
Problems in co-NP (maybe)
e.g., complement of partition problem, i.e., is there no
way to partition a set of numbers into 2 subsets with the same sum
Extensions of SAT (maybe)
Think of SAT as having a quantifier: do there exist x1,
x2, …, xn st Φ(x1,x2,…,xn) = true
You could add more quantifiers
For example, do there exist x1, x2, …,
xn st for all y1,…,ym
Φ(x1,x2,…xn,y1…,ym) = true
Limit of this process of adding quantifiers is “quantified
boolean formulas” which is PSPACE complete
PSPACE = problems that require polynomial memory
Example: construct 3-CNF formula from circuit
Part 1: circuit to formula reduction
You can’t just substitute formulas for gate inputs into formulas
for gates, because fan-out might grow exponentially
So construct formulas for values on wires
Part 2: general formula to 3-CNF formula reduction
Express formula as a tree of binary boolean operations; label
edges in tree with new variables
This gives formula as a conjunction of clauses with at most 3
literals in each
But those clauses aren’t necessarily disjunctions
Next
Finish exploring reduction from formula to 3-CNF
Then NP completeness in a wider variety of contexts