Spelling suggestions: "subject:"proposition proof complexity""
1 |
Expansion, Random Graphs and the Automatizability of ResolutionZabawa, Daniel Michael 25 July 2008 (has links)
We explore the relationships between the computational problem of recognizing expander graphs, and the problem of efficiently approximating proof length in the well-known system of \emph{resolution}. This program builds upon known connections between graph expansion and resolution lower bounds.
A proof system $P$ is \emph{(quasi-)automatizable} if there is a search algorithm which finds a $P$-proof of a given formula $f$ in time (quasi)polynomial in the length of a shortest $P$-proof of $f$. It is open whether resolution is (quasi-)automatizable. We prove several conditional non-automatizability results for resolution modulo new conjectures concerning the complexity of identifying bipartite expander graphs. Our reductions use a natural family of formulas and exploit the well-known relationships between expansion and length of resolution proofs. Our hardness assumptions are unsupported; we survey known results as progress towards establishing their plausibility. The major contribution is a conditional hardness result for the quasi-automatizability of resolution.
|
2 |
Expansion, Random Graphs and the Automatizability of ResolutionZabawa, Daniel Michael 25 July 2008 (has links)
We explore the relationships between the computational problem of recognizing expander graphs, and the problem of efficiently approximating proof length in the well-known system of \emph{resolution}. This program builds upon known connections between graph expansion and resolution lower bounds.
A proof system $P$ is \emph{(quasi-)automatizable} if there is a search algorithm which finds a $P$-proof of a given formula $f$ in time (quasi)polynomial in the length of a shortest $P$-proof of $f$. It is open whether resolution is (quasi-)automatizable. We prove several conditional non-automatizability results for resolution modulo new conjectures concerning the complexity of identifying bipartite expander graphs. Our reductions use a natural family of formulas and exploit the well-known relationships between expansion and length of resolution proofs. Our hardness assumptions are unsupported; we survey known results as progress towards establishing their plausibility. The major contribution is a conditional hardness result for the quasi-automatizability of resolution.
|
Page generated in 0.393 seconds