LARA

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
Next revision Both sides next revision
sav08:dpll_algorithm_for_sat [2008/03/12 17:34]
vkuncak
sav08:dpll_algorithm_for_sat [2008/03/12 19:31]
vkuncak
Line 72: Line 72:
 ===== Emiting a Satisfying Assignments or a Proof ===== ===== Emiting a Satisfying Assignments or a Proof =====
  
-Satisfying ​assignment: propagate or branch until map is complete.+Instead of returning just true/false we want SAT solver to return 
 +  * one satisfying ​assignment, if $S$ is satisfiable (easy: print S.A) 
 +  * proof, if $S$ is unsatisfiable - in some format
  
-Proof format.+Emiting resolution proofs - needs extra work
  
-===== DPLL Imperatively =====+**Unit resolution** is resolution, so we can construct proofs directly.
  
 +**Decision step** (picking variable value): from proofs for
 +\[
 +    S' \cup \{p\} \vdash {\it false}
 +\]
 +\[
 +    S' \cup \{\lnot p\} \vdash {\it false}
 +\]
 +we would like to construct the proof for
 +\[
 +    S' \vdash {\it false}
 +\]
 +++++Idea:|
 +From
 +\[
 +    S' \cup \{p\} \vdash {\it false}
 +\]
 +derive proof tree for
 +\[
 +    S' \vdash \lnot p
 +\]
 +and from
 +\[
 +    S' \cup \{\lnot p\} \vdash {\it false}
 +\]
 +derive proof tree for
 +\[
 +    S' \vdash p
 +\]
 +Then combine these two trees with resolution on $\{p\}$ and $\{\lnot p\}$ to get ${\it false}$.
  
-Selecting order of variables.+Why can we modify resolution proof to move $p$ from assumption and put its negation to conclusion?​ 
 +++++
  
-Lemma learning.+=== Lower Bounds on Running Time === 
 + 
 +Observation:​ constructed resolution proof is polynomial in the running time of the DPLL algorithm. 
 + 
 +Theorem: for some formulas, shortest resolution proofs are exponential. 
 + 
 +This does not contradict P vs NP question, because there may be "​better"​ proof systems than resolution. 
 + 
 +Lower bounds for both resolution and a stronger system are shown here by proving that interpolants can be exponential,​ and that interpolants are polynomial in proof size: 
 +  * Pavel Pudlák: [[http://​citeseer.ist.psu.edu/​36219.html|Lower Bounds for Resolution and Cutting Plane Proofs and Monotone Computations]]