Tatsuya Gima,Yuni Iwamasa,Yasuaki Kobayashi,Kazuhiro Kurita,Yota Otachi,Rin Saito
Abstract:In many decision-making processes, one may prefer multiple solutions to a single solution, which allows us to choose an appropriate solution from the set of promising solutions that are found by algorithms. Given this, finding a set of \emph{diverse} solutions plays an indispensable role in enhancing human decision-making. In this paper, we investigate the problem of finding diverse solutions of Satisfiability from the perspective of parameterized complexity with a particular focus on \emph{tractable} Boolean formulas. We present several parameterized tractable and intractable results for finding a diverse pair of satisfying assignments of a Boolean formula. In particular, we design an FPT algorithm for finding an ``almost disjoint'' pair of satisfying assignments of a $2$CNF formula.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: finding diverse pairs of solutions in the Boolean satisfiability problem (SAT), especially exploring how to effectively find two solutions with the maximum Hamming distance within the framework of parameterized complexity theory. Specifically, the authors focus on finding a pair of satisfying assignments with as large a difference as possible in a given Boolean formula, thereby enhancing the flexibility and diversity in the human decision - making process.
### Main contributions of the paper:
1. **Formalization of the diverse pair of solutions problem**:
- The authors formalize the problem as "Diverse Pair of Solutions" and "Dissimilar Pair of Solutions". The former requires finding two satisfying assignments such that the number of different variables between them is at least \(d\), and the latter requires that the number of different variables is at least \(|X|-s\), where \(X\) is the set of variables.
2. **Complexity analysis**:
- For different types of Boolean formulas (such as 2CNF, Horn, bi - Horn, and XOR formulas), the authors analyze the parameterized complexity of these problems. For example, for 2CNF formulas, Dissimilar Pair of Solutions is W[1] - hard, while for XOR formulas, Diverse Pair of Solutions is fixed - parameter tractable (FPT), and Dissimilar Pair of Solutions is W[1] - hard.
3. **Algorithm design**:
- The authors design fixed - parameter tractable algorithms for specific types of formulas (such as 2CNF formulas). For example, by reducing the problem to Almost 2SAT with hard constraints and using a vertex - cover - like algorithm in the literature to solve it.
4. **Handling of special cases**:
- For XOR formulas, the authors transform it into a problem of linear equations and prove that finding solutions with high Hamming weight is fixed - parameter tractable, while finding almost completely different pairs of solutions is W[1] - hard.
### Formula summary:
- **Hamming distance**: For two truth assignments \(\alpha_1\) and \(\alpha_2\), the Hamming distance is defined as:
\[
|\alpha_1\triangle\alpha_2|=\{x\in X:\alpha_1(x)\neq\alpha_2(x)\}
\]
- **Diverse Pair of Solutions**: Find two satisfying assignments \(\alpha_1,\alpha_2\) such that:
\[
|\alpha_1\triangle\alpha_2|\geq d
\]
- **Dissimilar Pair of Solutions**: Find two satisfying assignments \(\alpha_1,\alpha_2\) such that:
\[
|\alpha_1\triangle\alpha_2|\geq|X|-s
\]
### Conclusion:
This paper, through the parameterized complexity analysis of different types of Boolean formulas, provides effective algorithms or complexity results for finding diverse pairs of solutions in multiple situations, especially making significant progress on 2CNF and XOR formulas. These results not only deepen our understanding of the SAT problem but also provide theoretical support for finding diverse solutions in practical applications.