SAT v. SMT v. CP “Rat’s eye view”
Project Submissions
Scheduling meeting
SAT(-esque!) options:
Not this
One kind of project
NB: What you are /not/ implementing
You would not be implementing a SAT solver. Please. There are plenty of world-class implementations that are suuupper duper fast that have already been implemented.
Yours would suck by comparison.
You are implementing, if you will, a Sudoku solver. Or an 0h-h1 solver. Specifically, you implement it by implementing a (polynomial-time, many-one) reduction.
Karp’s paper on the subject
Many-one
For ` A <=_M B` you only get to invoke the solver of B, once, in the end.
Poly-time
The reduction must be polynomial in the size of the input, or else we haven’t really reduced the problem to the problem in question.
SAT : how?
DPLL
The main basis for solvers in this space. There are multiple advancements in this area, driven largely by heuristics and the fact that industrial, actual instances have more regularity and structure than random instances.
NP-/easy/
Other space: SAT v. SMT
Satisfiability, modulo theory.
Theory? Some interpreted theory with a specialized solver. Normally we find if something is true under any interpretation. Often we don’t care, and having to encode these behaviors as truth under all interpretations causes too much work. Instead we can rely on specialized solvers for given theories, under interpretations.
If you want a background-y “technical introduction to SMT” Video