Skip to content

Latest commit

 

History

History
36 lines (28 loc) · 1.72 KB

README.md

File metadata and controls

36 lines (28 loc) · 1.72 KB

Minimal models

Uses a SAT solver to find minimal partial assignments that are model of a CNF formula. Supports incremental clause additions. Quick and dirty implementation to illustrate the technique.

This works by maintaining two SAT solver instances, one for the CNF formula (the positive solver) and one for its negation (the negative solver). As the negation of a CNF formula is in DNF, we use the Tseytin transformation to convert the resulting DNF back into an equisatisfiable CNF with auxiliary variables.

To find a model we first query the positive solver and obtain a complete assignment that is a model of our formula (if it exists). If we query the negative solver using a model (complete or partial) of our formula as assumptions, the result will be UNSAT. The negative solver will then report a (not necessarily strict) subset of assumptions that are sufficient for falsifying the formula of the negative solver (called failed literals).

We can minimize the model by iteratively removing a literal from the assignment and querying the negative solver. If the result becomes SAT, we know the removed literal is essential and add it back to the assignment. If the result remains UNSAT, we can permanently remove that candidate and continue with another literal not known to be essential until none are left. Each time the result is UNSAT we can also remove all literals not among the failed literals.

I don't have anything to cite right now, but I'm pretty sure this is a known technique, very similar ideas are used all over the place as part of algorithms for MaxSAT, QBF, interpolant generation etc.

Usage

cat example.cnf - | cargo run --release

Then press return to find a minimal model and add a clause blocking the found model.