Graph Isomorphism, Tree Width, Path Width and LogSpace

Every once in a while, I can’t help thinking about “the complexity of graph isomorphism for bounded treewidth graphs“. Today has been one of those days again. See my earlier post to get the context.

Theorem ([Das, Toran and Wagner’10]) : Graph isomorphism of bounded treewidth graphs is in LogCFL.

The proof of the above theorem is as follows

  1. Graph isomorphism of bounded tree-distance width graphs is in L.
  2. Given two graphs and their tree decompositions, computing the isomorphism respecting these tree decompositions is reducible to (1).
  3. Given tree decomposition of only one graph, we can guess the tree decomposition of the other and guess the isomorphism (respecting the tree bags) and verify them using a non-deterministic auxiliary pushdown automata (a.k.a LogCFL).
  4. Since tree decomposition of a graph can be computed in LogCFL, the above theorem follows.
One of the bottlenecks, finding a tree decomposition of bounded treewidth graphs in logspace, is resolved by [Elberfeld, Jakoby and Tantau’10]. The following seems to be another major bottleneck.
Given a graph G and a decomposition D, how fast can we verify that D is a valid tree decomposition of G ? The upper bound of LogDCFL (the deterministic version of LogCFL) is clear from the above mentioned results. Can this verification be done in logspace ?
The answer is frustratingly unknown. An even more frustrating realization I had today is that “it is not clear how to beat the LogDCFL upper bound for the more restricted path decomposition“. Even though the underlying tree in a path decomposition is just a path, verifying the connectivity conditions of a path decomposition does seem to require recursion. It is not clear how to avoid recursion.
I thought that logspace upper bound is possible. Now I am much less confident about logspace upper bound. I cannot waste more time on this.
The truth is “this is a cute problem“. I need to do something to take my mind off this problem and move on. Easy enough, except I need an idea.
Update (Oct 12 2011) : Noticed that verification of path decompositions is easy.

Graph Isomorphism and Bounded Tree Width

If you read my earlier post, you known that I am fan of treewidth. Who isn’t !! The complexity of Graph Isomorphism (earlier post) is one of the long-standing open problem. Intersecting these two with one of my research interests (space-bounded computation) we get the following open problem :

Open Problem : What is the complexity of graph isomorphism for graphs with bounded treewidth ?

Graphs with treewidth at most k are also called partial k-trees. In 1992 Lindell proved that trees (graphs with treewidth=1) can be canonized in logspace [Lindel’92]. What about canonization for k=2 ? Recently Datta et. al [DLNTW’09] proved that the canonization of planar graph is logspace-complete. The following simple exercise shows that partial 2-trees are planar graphs. Hence the result of [DLNTW’09] implies that partial 2-trees can be canonized in logspace.

Exercise : Every partial 2-tree is planar.

In fact, canonization of partial 2-trees is settled earlier by [ADK’08]. What about k=3 ? Partial 3-trees may not be planar. An example is K_{3,3} itself. The tree width of K_{3,3} is three. I wanted to work on the case of k=3 and realized the following simple fact.

Exercise : Partial 3-trees are K_{5}-free.

In a follow-up paper to [DLNTW’09], Datta et al [DNTW’10] proved that canonization of K_{3,3}-free and K_{5}-free graphs is in Log-space. Hence we get the following corollary :

Corollary : Partial 3-trees can be canonized in log-space.

Since the above result is not explicitly mentioned in any papers, I wanted to make it clear in this post. Hence the open problem is for k \geq 4. LogCFL is the best known upper bound for graph isomorphism of partial k-trees [DTW’10]. One of the bottleneck, finding a tree decomposition of partial k-tree in logspace, is resolved recently [EJT’10]. The above mentioned papers make use of a decomposition of the input graph into two- or three-connected subgraphs, constructing an appropriate tree of these subgraphs, using the known structural properties of two- and three-connected graphs to canonize these subgraphs and using Lindell’s result to canonize the entire graph. Unfortunately no clean characterization exists for graphs with connectivity at least four. Many long-standing open problems in graph theory are trivial for 2- and 3- connected graphs and open for higher connectivity. A clean characterization of 4-connected graphs seems to be a major bottleneck in improving the space complexity of canonization of partial 4-trees. I am lost 😦

Open Problems

  • Is graph isomorphism of partial k-trees (for k \geq 4) in logspace ?
  • Is canonization of partial k-trees in LogCFL ? The paper of [DTW’10] solves isomorphism only.

References :

  • [Lindell’92] Steven Lindell: A Logspace Algorithm for Tree Canonization STOC 1992: pages 400-404
  • [DLNTW’09] Samir Datta, Nutan Limaye, Prajakta Nimbhorkar, Thomas Thierauf, Fabian Wagner: Planar Graph Isomorphism is in Log-Space. IEEE Conference on Computational Complexity 2009: 203-214
  • [DNTW’10] Samir Datta, Prajakta Nimbhorkar, Thomas Thierauf, Fabian Wagner: Graph Isomorphism for K{3, 3}-free and K5-free graphs is in Log-space. Electronic Colloquium on Computational Complexity (ECCC) 17: 50 (2010)
  • [ADK’08] Vikraman Arvind, Bireswar Das, Johannes Köbler: A Logspace Algorithm for Partial 2-Tree Canonization. CSR 2008: 40-51
  • [DTW’10] Bireswar Das, Jacobo Torán, Fabian Wagner: Restricted Space Algorithms for Isomorphism on Bounded Treewidth Graphs. STACS 2010: 227-238
  • [EJT’10] Michael Elberfeld, Andreas Jakoby, Till Tantau: Logspace Versions of the Theorems of Bodlaender and Courcelle. FOCS 2010: 143-152

Type Sensitive Depth and Karchmer Wigderson Games

Throughout this post, we will be considering circuits over the basis \{\vee,\wedge,\neg\} where \{\vee,\wedge\}-gates have fanin 2 and \neg-gates are only applied to input variables. Let f : \{0,1\}^n \rightarrow \{0,1\} be a boolean function on n variables and G_n be a circuit computing f. For an output gate g, let g_l and g_r be the sub-circuits, whose outputs are inputs to g. Let d(G_n) be the depth of circuit G_n and d(f) be the minimum depth of a circuit computing f.

Karchmer and Wigderson [KW’90] showed an equivalence between circuit depth and a related problem in communication complexity. It is a simple observation that we can designate the two players as an “and-player” and an “or-player”. Let S_0, S_1 \subseteq \{0,1\}^n such that S_0 \cap S_1 = \emptyset. Consider the communication game between two players (P_{\wedge} and P_{\vee}), where P_{\wedge} gets x \in S_1 and P_{\vee} gets y \in S_0. The goal of the players to find a coordinate i such that x_i \neq y_i. Let C(S_1,S_0) represent the minimum number of bits they have to communicate in order for both to agree on such coordinate.

Karchmer-Wigderson Theorem : For every function f : \{0,1\}^n \rightarrow \{0,1\} we have d(f) = C(f^{-1}(1),f^{-1}(0)).

Karchmer and Wigderson used the above theorem to prove that ‘monotone circuits for connectivity require super-logarithmic depth’. Let C_{\wedge}(S_1,S_0) (resp. C_{\vee}(S_1,S_0)) represent the minimum number of bits that P_{\wedge} (resp P_{\vee}) has to communicate. We can define type-sensitive depths of a circuit as follows. Let d_{\wedge}(G_n) (resp. d_{\vee}(G_n)) represent the AND-depth (resp. OR-depth) of G_n.

AND-depth : AND-depth of an input gate is defined to be zero. AND-depth of an AND gate g is max(d_{\wedge}(g_l), d_{\wedge}(g_r)) + 1. AND-depth of an OR gate g is max(d_{\wedge}(g_l), d_{\wedge}(g_l)). AND-depth of a circuit G_n is the AND-depth of its output gate.

OR-depth is defined analogously. Let d_{\wedge}(f) (resp. d_{\vee}(f)) be the minimum AND-depth (resp. OR-depth) of a circuit computing f.

Observation : For every function f : \{0,1\}^n \rightarrow \{0,1\} we have that C_{\wedge}(f^{-1}(1),f^{-1}(0)) corresponds to the AND-depth and C_{\vee}(f^{-1}(1),f^{-1}(0)) corresponds to the OR-depth of the circuit constructed by Karchmer-Wigderson.


Open Problems :

  • Can we prove explicit non-trivial lower bounds of d_{\wedge}(f) (or d_{\vee}(f)) of a given function f ? This sort of “asymmetric” communication complexity is partially addressed in [MNSW’98].
  • A suitable notion of uniformity in communication games is to be defined to address such lower bounds. More on this in future posts.


References :

  • [KW’90] Mauricio Karchmer and Avi Wigderson : Monotone circuits for connectivity require super-logarithmic depth. SIAM Journal on Discrete Mathematics, 3(2):255–265, 1990.
  • [MNSW’98] Peter Bro Miltersen, Noam Nisan, Shmuel Safra, Avi Wigderson: On Data Structures and Asymmetric Communication Complexity. J. Comput. Syst. Sci. 57(1): 37-49 (1998)


Balanced ST-Connectivity

Today’s post is about a new open problem arising from my recent paper  (available on ECCC). The problem is as follows :

Let G(V,E) be a directed graph. Let G'(V,E') be the underlying undirected graph of G. Let P be a path in G'. Let e = (u,v) be an edge along the path P. Edge e is called neutral edge if both (u,v) and (v,u) are in E. Edge e is called forward edge if (u,v) \in E and (v,u) \notin E. Edge e is called backward edge if (u,v) \notin E and (v,u) \in E.

A path (say P) from s \in V to t \in V in G'(V,E') is called balanced if the number of forward edges along P is equal to the number of backward edges along P. A balanced path might have any number of neutral edges. By definition, if there is a balanced path from s to t then there is a balanced path from t to s. The path P may not be a simple path. We are concerned with balanced paths of length at most n.

Balanced ST-Connectivity : Given a directed graph G(V,E) and two distinguished nodes s and t, decide if there is balanced path (of length at most n) between s and t.

In my paper, I proved that SGSLOGCFL, a generalization of Balanced ST-Connectivity, is contained in DSPACE(lognloglogn). Details about SGSLOGCFL are in my paper.

Theorem 1 : SGSLOGCFL is in DSPACE(lognloglogn).

Open Problem : Is SGSLOGCFL \in L ?

Cash Prize : I will offer $100 for a proof of SGSLOGCFL \in L. I have spent enough sleepless nights trying to prove it. In fact, an alternate proof of Theorem 1 (or even any upper bound better than O({\log}^2n)) using zig-zag graph product seems to be a challenging task.

Usually people offer cash prizes for a mathematical problem when they are convinced that :

  • it is a hard problem.
  • it is an important problem worth advertising.
  • the solution would be beautiful, requires new techniques and sheds new light on our understanding of related problems.

My reason is “All the above”. Have Fun solving it !!

A cute puzzle : In Balanced ST-Connectivity we are only looking for paths of length at most n. There are directed graphs where the only balanced st-path is super-linear. The example in the following figure shows an instance of Balanced ST-Connectivity where the only balanced path between s and t is of length \Theta(n^2). The directed simple path from s to t is of length n/2. There is a cycle of length n/2 at the vertex v. All the edges (except (v,u)) on this cycle are undirected. The balanced path from s to t is obtained by traversing from s to v, traversing the cycle clockwise for n/2 times and then traversing from v to t.

Puzzle : Are there directed graphs where every balanced st-path is of super-polynomial size ?

Update : The above puzzle is now solved.

Open Problems

  • Is SGSLOGCFL \in L ?
  • Are there directed graphs where every balanced st-path is of super-polynomial size ? (solved)
  • More open problems are mentioned in my paper.

Hardness of Graph Isomorphism

The complexity of Graph Isomorphism (GI) is one of the major open problems. It is easy to see that GI \in NP. It is known that GI \in NP \cap coAM. The following theorem states that it is unlikely that GI is NP-complete.

Theorem [Schöning’87, BHZ’87] : If GI is NP-complete then the polynomial hierarchy collapses to its second level.

The counting version of GI is known to be reducible to its decisional version. A polynomial time algorithm solving GI would be a major breakthrough. The best known algorithm runs in 2^{O(\sqrt{n{\log}n})} for graphs with n vertices. Several special cases are shown to be in P. Several problems are known to be GI-hard. See this wikipedia article for details. GI is widely believed to be an NP-intermediate problem.

Conjecture : If P \neq NP, then GI is neither NP-complete nor in P.

Note that if the above conjecture is true then GI is P-hard. Is GI known to be P-hard ? What is the best known hardness of GI ? Well… we know very little about the hardness of GI. The following exercises show that GI is L-hard.

Exercise : Consider the following restricted automorphism problem: Given a graph G = (V,E) and two lists of nodes (x_1, \dots, x_k),(y_1,\dots, y_k), is there an automorphism in G mapping x_i to y_i for 1 ≤ i ≤ k ? Show that this problem is reducible to GI.

Exercise : Show that Undirected ST-connectivity is reducible to the above mentioned automorphism problem.

Torán [Torán’00] proved the following hardness theorem. Informally speaking, GI is hard for all complexity classes defined in terms of the number of accepting computations of a nondeterministic logarithmic space machine. These are the best known hardness results for GI.

Theorem [Torán’00] : GI is hard for NL, PL, Mod_k{L} and DET.

All these hardness results are under DLOGTIME uniform AC^0 many-one reductions. DET is the class of problems NC^1 Turing reducible to the determinant [Cook’85]. It is known that Mod_k{L} \subseteq DET and NL \subseteq C_{=}L \subseteq PL \subseteq DET. Hence the best known hardness of GI is DET-hardness. However, we do not know the exact complexity of DET i.e., we don’t know where DET lies in terms of the known complexity classes between NL and NC^2. In particular, what is the relation between LogCFL = SAC^1 and DET ?

Torán also showed a randomized logarithmic space reduction from the perfect matching problem to graph isomorphism. More details about the complexity of perfect matching in a future blog post.

Open Problems:

  • Is GI LogCFL-hard ?
  • Is DET LogCFL-hard ? What is the relation between LogCFL and DET ? This is an independent long-standing open problem. It deserves a separate blog post.
  • Is GI \in coNP ? A proof of this would imply that “if GI is NP-complete then NP = coNP“, improving the above mentioned theorem.
  • Is GI in P for strongly regular graphs ? The best known algorithm for strongly regular graphs, given by Spielman [Spielman’96], runs in time n^{O({n^{1/3}}{{\log}n})}.

References :

  • [BHZ’87] R. Boppana, J. Håstad, and S. Zachos , “Does co-NP have short interactive proofs?”, Information Processing Letters 25(2), pages 127-132, (1987).
  • [Schöning’87] Uwe Schöning, Graph isomorphism is in the low hierarchy, Proceedings of the 4th Annual Symposium on Theoretical Aspects of Computer Science, 1987, 114–124; also: Journal of Computer and System Sciences, vol. 37 (1988), 312–323
  • [Cook’85] Stephen A. Cook, A Taxonomy of Problems with Fast Parallel Algorithms Information and Control 64(1-3): 2-21 (1985)
  • [Spielman’96] Daniel A. Spielman, Faster isomorphism testing of strongly regular graphsSTOC ’96: Proceedings of the twenty-eighth annual ACM symposium on Theory of computing, ACM, pp. 576–584
  • [Torán’00] Jacobo Torán, On the Hardness of Graph Isomorphism. FOCS’2000, also: SIAM J. Comput. 33(5): 1093-1108 (2004)

NP intersect coNP

NP is the set of languages that have short proofs. coNP is the set of languages that have short refutations. Note that coNP is not the complement of NP. NP \cap coNP is non-empty. It is easy to see that all languages in P are in NP \cap coNP i.e., P \subseteq NP \cap coNP. It is conjectured that P \subsetneq NP \cap coNP. i.e., there are problems in NP \cap coNP that are not in P. Following are some problems in NP \cap coNP that are not known to be in P.

Factoring : Given an integer what is the complexity of finding its factors. Every integer always has a unique factorization. Hence, Factoring is very different from the NP-complete problems. The following exercise states that it is highly unlikely that Factoring is NP-complete. On the other hand if Factoring is in P then the world as we know today will be in chaos !! Factoring is conjectured to be an intermediate problem.

Exercise : If Factoring is NP-complete then NP = coNP.

The first step to solve the above exercise is to show that Factoring is in NP \cap coNP. In fact it is also in UP \cap coUP. Perhaps this is the strongest evidence that P \subsetneq NP \cap coNP.

Parity Games : Deciding which of the two players has a winning strategy in parity games is in NP \cap coNP, as well as in UP \cap coUP.

Stochastic Games : The problem of deciding which player has the greatest chance of winning a stochastic game is in NP \cap coNP [Condon’92]

Lattice Problems : The problems of approximating the shortest and closest vector in a lattice to within a factor of \sqrt{n} is in NP \cap coNP [AR’05].

All the above problems are not known to be in P.

Open Problems :

  • Are there other problems in NP \cap coNP that are not known to be in P.
  • PPAD, PLS have been defined to understand problems whose structure is different from NP-complete problems. Can we define new complexity classes to study the complexity of the above mentioned problems (and related problems if any) ?
  • Graph Isomorphism (GI) is also conjectured to be an intermediate problem. It is known that GI is not NP-complete unless Polynomial Hierarchy collapses to its second level. Can we improve this result by showing that GI is in coNP ? Whether GI is in coNP is an interesting open problem for a very different reason also. More on this in a future post.

References :

  • [Condon’92] Anne Condon: The Complexity of Stochastic Games Inf. Comput. 96(2): 203-224 (1992)
  • [AR’05] Dorit Aharonov, Oded Regev: Lattice problems in NP \cap coNP. J. ACM 52(5): 749-765 (2005)

Logspace vs Polynomial time

One of the primary goals of complexity theory is separating complexity classes, a.k.a proving lower bounds. Embarrassingly we have only a handful of unconditional separation results. Separating P from NP is of course the mother of all such goals. Anybody who understands the philosophical underpinnings of the P vs NP problem would love to LIVE to see its resolution. Towards resolving this, we made some (“anti”)-progress (Eg : Relativization, Natural proofs, Algebrization) and have a new geometric complexity theory approach which relies on an Extended-Extended-Extended-Extended-Riemann-Hypothesis !! For more information about the history and status of P vs NP problem read Sipser’s paper [Sipser’92], Allender’s status report [Allender’09] or Fortnow’s article [Fortnow’09].

Today’s post is about the Logspace (L) vs Polynomial time (P) problem, which (in my opinion) is right next to the P vs NP problem in its theoretical importance. I guess many researchers believe that L \neq P. Did we make any progress/anti-progress towards resolving the L \neq P conjecture ?  Here are two attempts both based on branching programs and appeared in MFCS with a gap of 20 years !!

1) A conjecture by Barrington and McKenzie (BM’89): The problem GEN is defined as follows :

GEN : Given an n \times n table filled with entries from \{1,2,\dots,n\}, which we interpret as the multiplication table of an n-element groupoid, and a subset S of \{1,2,\dots,n\} which includes element 1, determine whether the subgroupoid <S>, defined as the closure of S under the groupoid product, includes element n.

Barrington-McKenzie Conjecture : For each n > 1, a branching program in which each node can only evaluate a binary product within an n-element groupoid, branching n ways according to the n possible outcomes, must have at least 2^{n-2} nodes to solve all n \times n GEN instances with singleton starting set S.

The problem GEN is known to be P-complete [JL’76]. Barrington-McKenzie Conjecture would imply that GEN \notin DSPACE({{\log}^k}n) for any k. In particular, it would imply that L \neq P. I don’t know if there is any partial progress towards resolving this conjecture.

2) Thrifty Hypothesis : This is a recent approach by Braverman et. al [BCMSW’09] towards proving a stronger theorem L \neq LogDCFL. Stephen Cook presented this approach at Valiant’s 60th birthday celebration and Barriers Workshop. He also announced a $100 prize for solving an intermediate open problem mentioned in his slides.

Tree Evaluation Problem (TEP): The input to the problem is a rooted, balanced d-ary tree of height h, whose internal nodes are labeled with d-ary functions on [k] = \{1, . . . , k\}, and whose leaves are labeled with elements of [k]. Each node obtains a value in [k] equal to its d-ary function applied to the values of its d children. The output is the value of the root.

In their paper they show that TEP \in LogDCFL and conjecture that TEP \notin L. They introduce Thrifty Branching Programs and prove that TEP can be solved by a thrifty branching program. A proof of the following conjecture implies that L \neq LogDCFL. For more details, read this paper.
Thrifty Hypothesis : Thrifty Branching Programs are optimal among deterministic branching programs solving TEP.

Open Problems :

  • My knowledge about the history of L vs P problem is limited.  Are there other approaches/attempts in the last four decades to separate L from P ?
  • An intermediate open problem is mentioned in the last slide of these slides. The authors announced $100 prize for the first correct proof. Read their paper for more open problems.

References :

  • [BM’89] David A. Mix Barrington, Pierre McKenzie: Oracle Branching Programs and Logspace versus P. MFCS 1989: 370-379
  • [BCMSW’09] Mark Braverman, Stephen A. Cook, Pierre McKenzie, Rahul Santhanam, Dustin Wehr: Branching Programs for Tree Evaluation. MFCS 2009: 175-186
  • [Sipser’92] Michael Sipser: The History and Status of the P versus NP Question STOC 1992: 603-618
  • [Allender’09] Eric Allender: A Status Report on the P Versus NP Question. Advances in Computers 77: 117-147 (2009) [pdf]
  • [Fortnow’09] Lance Fortnow: The status of the P versus NP problem. Commun. ACM 52(9): 78-86 (2009) [pdf]
  • [JL’76] Neil D. Jones, William T. Laaser: Complete Problems for Deterministic Polynomial Time. Theor. Comput. Sci. 3(1): 105-117 (1976)