On Approximate Reconfigurability of Label Cover
Abstract
Given a two-prover game and its two satisfying labelings and , the Label Cover Reconfiguration problem asks whether can be transformed into by repeatedly changing the value of a vertex while preserving any intermediate labeling satisfying . We consider an optimization variant of Label Cover Reconfiguration by relaxing the feasibility of labelings, referred to as Maxmin Label Cover Reconfiguration: we are allowed to transform by passing through any non-satisfying labelings, but required to maximize the minimum fraction of satisfied edges during transformation from to . Since the parallel repetition theorem of RazΒ (SIAMΒ J.Β Comput.,Β 1998)Β [Raz98], which implies NP-hardness of Label Cover within any constant factor, produces strong inapproximability results for many NP-hard problems, one may think of using Maxmin Label Cover Reconfiguration to derive inapproximability results for reconfiguration problems. We prove the following results on Maxmin Label Cover Reconfiguration, which display different trends from those of Label Cover and the parallel repetition theorem:
-
β’
Maxmin Label Cover Reconfiguration can be approximated within a factor of nearly for restricted graph classes, including slightly dense graphs and balanced bipartite graphs.
-
β’
A naive parallel repetition of Maxmin Label Cover Reconfiguration does not decrease the optimal objective value.
-
β’
Label Cover Reconfiguration on projection games can be decided in polynomial time.
The above results suggest that a reconfiguration analogue of the parallel repetition theorem is unlikely.
1 Introduction
Background.
In reconfiguration problems [IDH+11], given a pair of feasible solutions for a combinatorial problem, we wish to find a step-by-step transformation from one to the other while the feasibility of every intermediate solution is maintained. Under the framework of reconfiguration [IDH+11], numerous reconfiguration problems have been derived from classical search problems; refer to the surveys by NishimuraΒ [Nis18] and van den HeuvelΒ [vdH13] for an overview of reconfiguration.
In this article, we consider reconfigurability of Label Cover [ABSS97] and its approximation. Given a two-prover game and its two satisfying labelings and , the Label Cover Reconfiguration problem asks whether can be transformed into by repeatedly changing the value of a vertex preserving any intermediate labeling satisfying . We can consider an optimization variant of this problem by relaxing the feasibility of labelings, referred to as Maxmin Label Cover Reconfiguration: we are allowed to transform by passing through any non-satisfying labelings, but required to maximize the minimum fraction of satisfied edges during transformation from to . Solving Maxmin Label Cover Reconfiguration approximately, we may find approximate reconfiguration for Label Cover Reconfiguration consisting of almost-satisfying labelings.
The recent work of the authorΒ [Ohs23] demonstrated that assuming PSPACE-hardness of approximation for Maxmin CSP Reconfiguration (which implies that for Maxmin Label Cover Reconfiguration), a bunch of reconfiguration problems are also PSPACE-hard to approximate, say, within a factor of for some . One limitation of this approach is that the value of such might be too small to rule out a -approximation algorithm. In NP optimization problems, the parallel repetition theorem due to RazΒ [Raz98] serves as the βmotherβ of many strong inapproximability results [Fei98, HΓ₯s99, HΓ₯s01]. The crux of this theorem is to imply along with the PCP theoremΒ [AS98, ALM+98] that for every , there is a finite alphabet such that Label Cover on is NP-hard to approximate within a factor of . One might thus think of a reconfiguration analogue of the parallel repetition theorem for Maxmin Label Cover Reconfiguration, which would help in improving PSPACE-hardness of approximation for reconfiguration problems. Our contribution is to give evidence that such hopes are probably dashed.
Our Results.
We present a few results on Maxmin Label Cover Reconfiguration, which display different trends from those for Label Cover and the parallel repetition theorem.
-
β’
In SectionΒ 2, we prove that Maxmin Label Cover Reconfiguration can be approximated within a factor of nearly for restricted graph classes, including slightly dense graphs and balanced bipartite graphs,
-
β’
In SectionΒ 3, we further show that a naive parallel repetition of Maxmin Label Cover Reconfiguration does not decrease the optimal objective value, defined as the worst fraction of satisfied edges during transformation.
-
β’
In SectionΒ 4, we develop a polynomial-time algorithm for deciding Label Cover Reconfiguration on projection games, which is based on a simple characterization of the reconfigurability between a pair of satisfying labelings.
Our results suggest that a reconfiguration analogue of the parallel repetition theorem is unlikely; thus, we should resort to a different approach to derive an improved factor of inapproximability for reconfiguration problems.
Preliminaries.
For two integers with , let and . For a statement , is if is true, and otherwise. We formally define Label Cover Reconfiguration and its optimization variant.A constraint graph is defined as a tuple , where
-
β’
is an undirected graph called the underlying graph of ,
-
β’
is a finite set called the alphabet, and
-
β’
is a collection of binary constraints, where each consists of pairs of admissible values that endpoints of can take.
If the underlying graph of is a bipartite graph, is called a two-prover game or simply game. We write to stress that the underlying graph of is a bipartite graph with bipartition . Unless otherwise specified, we use to represent a game. Moreover, is said to be a projection game if every constraint for has a projection property; i.e., each value for has a unique value for such that . Such is denoted .
A labeling for a constraint graph is a mapping that assigns value of to vertex of . We say that satisfies edge (or constraint ) if , and satisfies if it satisfies all edges of . For two satisfying labelings and for , a reconfiguration sequence from to is any sequence of labelings starting from and ending with such that each labeling is obtained from the previous one by changing the value of a single vertex; i.e., they differ in exactly one vertex. The Label Cover Reconfiguration problem is defined as follows:
[l]Label Cover Reconfiguration Input: a satisfiable game and its two satisfying labelings and . Question: is there a reconfiguration sequence of satisfying labelings from to ?
Label Cover Reconfiguration is known to be PSPACE-complete, e.g., [GKMP09]. We then proceed to an optimization variant [IDH+11] of Label Cover Reconfiguration, in which we are allowed to touch non-satisfying labelings. For a game and a labeling , let denote the fraction of edges satisfied by ; namely,
(1.1) |
For any reconfiguration sequence , let denote the minimum fraction of satisfied edges over all βs in ; namely,
(1.2) |
Then, Maxmin Label Cover Reconfiguration is defined as the following optimization problem:
[l]Maxmin Label Cover Reconfiguration Input: a satisfiable game and its two satisfying labelings and . Question: maximize subject to that is a reconfiguration sequence from to .
2 Nearly -factor Approximation
We show that Maxmin Label Cover Reconfiguration is approximately reconfigurable within a factor of nearly for some graph classes. For a game and its two labelings , let denote the maximum value of over all possible reconfiguration sequences from to ; namely,
(2.1) |
Theorem 2.1.
For a game having edges and two of its satisfying labelings , the following holds:
-
β’
If the average degree of is at least for , then . Moreover, the same result holds even if is a general (i.e., non-bipartite) constraint graph.
-
β’
If is balanced (i.e., the two parts have the same size), then .
The proof of TheoremΒ 2.1 relies on the following claim.
Claim 2.2.
For a constraint graph and its two satisfying labelings and , it holds that
(2.2) |
where is the edge set of a subgraph of induced by vertex set .
Proof.
For a vertex set , we define a labeling as follows:
(2.3) |
Consider transforming into by changing the value of from to one by one. Since the value of vertices in has never been changed, any edge of are always satisfied during this transformation; i.e., . Similarly, it follows that . Consequently, we derive
(2.4) |
as desired. β
We focus on the case that is balanced; i.e., . We wish to partition each and in a particular well-balanced manner. To this end, we use the following auxiliary lemma.
Lemma 2.3.
For any positive integers in the range of , there exists a partition of such that
(2.5) |
Moreover, such and can be found in polynomial time.
Proof.
The case of is trivial because and so ; we thus assume that . Denote for any . Consider the following naive greedy algorithm for Bin Packing:
[l]Greedy algorithm
We show
(2.6) |
Define
(2.7) |
Since for every , are at least and are at least , we have
(2.8) |
Observing easily that
(2.9) |
due to the nature of the greedy algorithm, we consider the following two cases:
- (Case 1)
-
If the absolute difference between and is larger than , then we would have added all of into either or . Thus, the absolute difference between and will be simply
(2.10) - (Case 2)
-
Otherwise, the absolute difference between and must be at most . Using Eq.Β 2.8, we can derive
(2.11) where the last inequality is due to the fact that .
In either case, we obtain
(2.12) |
Consequently, we derive
(2.13) |
completing the proof. β
We are now ready to prove TheoremΒ 2.1.
Proof of TheoremΒ 2.1.
By ClaimΒ 2.2, it suffices to show the existence of a desired partition of for each case. The statement for the case of large average degree directly follows from KΓΌhn and OsthusΒ [KO03, Corollary 18], stating that for every graph of edges and average degree , its vertex set can be partitioned into and such that both and contain at least edges.
Suppose then that is a balanced game over edges such that . Using LemmaΒ 2.3, we construct a pair of partitions and of and , respectively, such that
(2.14) |
where denotes the degree of in . Define . By case analysis, we show that either of or gives us a desired partition.
- (Case 1)
-
: and can be bounded from below as follows:
(2.15) (2.16) - (Case 2)
-
: Similarly to (Case 1), we can bound
(2.17) - (Case 3)
-
and : We are done since and have a desired property.
Consequently, in either case, we obtain
(2.18) |
which completes the proof. β
3 Parallel Repetition Does Not Amplify Gap
Here, we show that a naive parallel repetition of Label Cover Reconfiguration does not amplify the gap unlike the parallel repetition theorem [Raz98]. We first define the product of two games.
Definition 3.1.
Let and be two games. Then, the product of and , denoted , is defined as a new game , where
(3.1) |
and the constraint for each edge is defined as
(3.2) |
The -fold parallel repetition of , denoted , for any positive integer is defined as
(3.3) |
The parallel repetition theorem due to RazΒ [Raz98] states that for every game with , it holds that , where depends only on the value of .
We can think of a reconfiguration analogue for parallel repetition. More precisely, for two labelings for and for , the product labeling of and , denoted , is defined as a labeling such that
(3.4) |
We write for denoting It is easy to see that implies . So, given that , does the value of the -fold parallel repetition of a game decreases as the increase of ? Unfortunately, the answer is negative:
Observation 3.2.
Let be a satisfiable game, and and be its two satisfying labelings. Then, for every positive integer , it holds that
(3.5) |
Proof.
Let be a satisfiable game. Suppose that we are given a reconfiguration sequence from to such that . We will show
(3.6) |
To this end, for each , we bound
(3.7) |
Fix and . We assume that and differ in some ; the case that they differ in some can be shown similarly. Construct then a reconfiguration sequence from to by changing the value of a vertex of if its value is different between the two assignments. Obviously, it holds that
(3.8) |
Observe now that and differ on vertex if and only if and agree with each other on any vertex . Thus, for any assignment ,
for all and . By abuse of notation, denoting for assignment and edge ; we bound the value of as follows:
Since satisfies if and satisfies if , we further have
Concatenating reconfiguration sequences for all , we have
(3.9) |
Consequently, we obtain
(3.10) |
completing the proof. β
4 Polynomial-time Reconfiguration for Projection Games
We finally present a simple polynomial-time algorithm that decides reconfigurability between a pair of satisfying labelings for a projection game.
Theorem 4.1.
For a satisfiable projection game and its two satisfying labelings and , we can decide in polynomial time if there is a reconfiguration sequence from to of satisfying labelings for .
Proof.
Since each connected component of can be processed independently, we only consider the case that is connected. If consists of a single vertex (i.e., or is empty), the answer is always βreconfigurableβ as it has no edges; thus we can safely assume that . We show that and are reconfigurable to each other if and only if and agree on .
We first show the only-if direction. Suppose that for some . Starting from , changing the value of from to any other value must violate edges incident to since the value of should be mapped to via , as claimed.
We then show the if direction. Suppose that for every . Then, it holds that for every edge . We can thus obtain a transformation from to without breaking any constraint, by changing the value of each vertex of from to one by one, as desired. β
References
- [ABSS97] Sanjeev Arora, LΓ‘szlΓ³ Babai, Jacques Stern, and Z.Β Sweedyk. The hardness of approximate optima in lattices, codes, and systems of linear equations. J. Comput. Syst. Sci., 54(2):317β331, 1997.
- [ALM+98] Sanjeev Arora, Carsten Lund, Rajeev Motwani, Madhu Sudan, and Mario Szegedy. Proof verification and the hardness of approximation problems. J. ACM, 45(3):501β555, 1998.
- [AS98] Sanjeev Arora and Shmuel Safra. Probabilistic checking of proofs: A new characterization of NP. J. ACM, 45(1):70β122, 1998.
- [Fei98] Uriel Feige. A threshold of for approximating set cover. J. ACM, 45(4):634β652, 1998.
- [GKMP09] Parikshit Gopalan, PhokionΒ G. Kolaitis, Elitza Maneva, and ChristosΒ H. Papadimitriou. The connectivity of Boolean satisfiability: Computational and structural dichotomies. SIAM J. Comput., 38(6):2330β2355, 2009.
- [HΓ₯s99] Johan HΓ₯stad. Clique is hard to approximate within . Acta Math., 182:105β142, 1999.
- [HΓ₯s01] Johan HΓ₯stad. Some optimal inapproximability results. J. ACM, 48(4):798β859, 2001.
- [IDH+11] Takehiro Ito, ErikΒ D. Demaine, Nicholas J.Β A. Harvey, ChristosΒ H. Papadimitriou, Martha Sideri, Ryuhei Uehara, and Yushi Uno. On the complexity of reconfiguration problems. Theor. Comput. Sci., 412(12-14):1054β1065, 2011.
- [KO03] Daniela KΓΌhn and Deryk Osthus. Partitions of graphs with high minimum degree or connectivity. J. Comb. Theory. Ser. B, 88(1):29β43, 2003.
- [Nis18] Naomi Nishimura. Introduction to reconfiguration. Algorithms, 11(4):52, 2018.
- [Ohs23] Naoto Ohsaka. Gap preserving reductions between reconfiguration problems. In STACS, pages 49:1β49:18, 2023.
- [Raz98] Ran Raz. A parallel repetition theorem. SIAM J. Comput., 27(3):763β803, 1998.
- [vdH13] Jan vanΒ den Heuvel. The complexity of change. In Surveys in Combinatorics 2013, volume 409, pages 127β160. Cambridge University Press, 2013.