Download presentation
Presentation is loading. Please wait.
Published byDerek Bates Modified over 9 years ago
1
Project proposals
2
Most involve implementation inside minisat+ Project includes: Reading Implementation Evaluation Short presentation Submission of results
3
#0) Build your own SAT solver Well, with some assistance…
4
#1) Direct treatment of cardinality constraints in Minisat+ Today: reduction to CNF Tomorrow: direct treatment of cardinality constraints.
5
#2) incremental Today: monolithic formula Tomorrow: Add clauses incrementally. Wait to see which clauses are more active. Deactivate the rest, letting the solver focus on part of the formula. With partial / full assignment, check the rest. Add the clauses that contradict the assignment and restart.
6
#3) Parallelization Minisat+ is a reduction-based PBS solver. Today: Single-threaded Tomorrow: Multi-threaded Small differences in solvers (e.g., parameters, random seed) Sharing small clauses
7
#4) Add “Glue” information to minisat Today: importance of clauses is measured by their activities Tomorrow: Add information about the # of decision levels in the clause. Use it to invent deletion strategies. Use it for freezing/unfreezing conflict clauses Use it as a criterion for clause shrinking Predicting Learnt Clauses Quality in Modern SAT Solvers, Gilles Audemard, Laurent Simon, IJCAI09
8
#4) Freeze / Unfreeze conflict clauses Today: Tomorrow: Clauses Frozen clauses LearningForgetting Learning Freezing Unfreezing Forgetting On Freezing and Reactivating Learnt Clauses. Gilles Audemard, Jean-Marie Lagniez, Bertrand Mazure, and Lakhdar Saıs, SAT’11 + ‘Glue’ information
9
#5) BCP order Today: BCP order is arbitrary Tomorrow: smarter heuristics: e.g., Variables with higher activity score BFS / DFS among variables
10
#6) Modeling… Model an interesting problem … Examples from previous years: Class scheduling Exam scheduling Assignment to dorms Games…
11
#7) model class scheduling with difference constraints Today: class scheduling translates difference constraints to propositional logic. Tomorrow: Generate a formula in Integer Difference Logic and solve with an SMT-solver (e.g. Microsoft’s “Z3”).
12
#8) logarithmic encoding of constraints In the class scheduling system Today: direct (i.e., linear in the number of values) encoding Tomorrow: logarithmic encoding. The Log-Support Encoding of CSP into SAT/Marco Gavanelli
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.