Abstract:With the goal of obtaining strong relaxations for binary polynomial optimization problems, we introduce the pseudo-Boolean polytope defined as the convex hull of the set of binary points satisfying a collection of equations containing pseudo-Boolean functions. By representing the pseudo-Boolean polytope via a signed hypergraph, we obtain sufficient conditions under which this polytope has a polynomial-size extended formulation. Our new framework unifies and extends all prior results on the existence of polynomial-size extended formulations for the convex hull of the feasible region of binary polynomial optimization problems of degree at least three.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the strong relaxation representation of the pseudo - Boolean polytope in binary polynomial optimization problems. Specifically, the author introduced the concept of the pseudo - Boolean polytope and, through the signed hypergraph representation, studied under what conditions this polytope can have a polynomial - sized extended formulation. This research aims to provide a stronger linear programming relaxation form for high - degree binary polynomial optimization problems, thereby improving the solution efficiency.
### Main contributions of the paper
1. **Generalize existing results**:
- The author shows how to apply the hypergraph framework pioneered in multilinear optimization to pseudo - Boolean optimization and proposes sufficient conditions for the pseudo - Boolean polytope to have a polynomial - sized extended formulation. These results unify and extend all previous results on polynomial - sized representations of multilinear polytopes.
2. **Propose a new technical framework**:
- The author introduced a technical framework called "recursive inflate - and - decompose", which is used to construct polynomial - sized extended formulations of the pseudo - Boolean polytope. This framework relies on the recursive application of three key steps:
1. **Decomposability conditions of the pseudo - Boolean polytope** (see Theorem 1): This is the first result on the decomposability of the pseudo - Boolean polytope and is an important generalization of Theorem 4 in [20].
2. **Polynomial - sized extended formulation of the pseudo - Boolean polytope of pointed signed hypergraphs** (see Theorem 2): This is the basic module for constructing the extended formulation.
3. **Inflation operation of signed edges** (see Theorem 3): Through this operation, a large class of signed hypergraphs can be transformed into a form suitable for the previous two partial results.
3. **Specific results**:
- If the underlying hypergraph of the signed hypergraph \( H \) is β - acyclic, then \( P_{pB}(H) \) has a polynomial - sized extended formulation (see Theorem 5).
- If the underlying hypergraph of the signed hypergraph \( H \) is α - acyclic and has a log - polynomial rank, then \( P_{pB}(H) \) has a polynomial - sized extended formulation (see Theorem 6).
- The concept of "gap" is introduced to efficiently inflate signed edges, and combined with the results of Theorem 3 and Theorem 5, more polynomial - sized extended formulations of the pseudo - Boolean polytope are obtained in more cases (see Proposition 3 and Proposition 4).
### Technical details
1. **Definition of the pseudo - Boolean polytope**:
- The pseudo - Boolean polytope \( P_{pB}(H) \) is the convex hull of the pseudo - Boolean set \( S_{pB}(H) \) in the signed hypergraph \( H=(V, S) \), where \( S_{pB}(H) \) is defined as all \( z\in\{0, 1\}^{V\cup S} \) that satisfy \( z_s = \prod_{v\in s}\sigma_s(z_v) \).
2. **Decomposability conditions**:
- The pseudo - Boolean polytope \( P_{pB}(H) \) can be decomposed into \( P_{pB}(H_1) \) and \( P_{pB}(H_2) \) if the following conditions are met:
- \( H_1 \) and \( H_2 \) are sub - signed hypergraphs constructed from \( H \) in a specific way.
- If \((z_1, z_{\cap})\in P_{pB}(H_1) \) and \((z_{\cap}, z_2)\in P_{pB}(H_2) \), then \((z_1, z_{\cap}, z_2)\in P_{pB}(H) \).
3. **Inflation operation**:
- By inflating signed edges, complex signed hypergraphs can be simplified into simpler forms, making it easier to construct polynomial - sized extended formulations.
### Conclusion
This paper significantly advances the research on the pseudo - Boolean polytope by introducing new techniques and frameworks, providing a stronger linear programming relaxation form for high - degree binary polynomial optimization problems. These results are not only of great theoretical significance but also have potential value in practical applications.