P-complete

Alternating Graph Accessibility Problem (AGAP)

Category: Graph Theory

Input

A directed graph $G = (V, E)$, a partition $V = A \cup B$ of the vertices, and designated vertices $s$ and $t$.

Problem

Is $\text{apath}(s, t)$ true?

Definitions

The predicate $\text{apath}$ is defined as follows. Vertices in $A$ are “universal,” those in $B$ are “existential.” Such a graph is called an alternating graph or an AND/OR graph. The predicate $\text{apath}(x, y)$ holds if and only if:

  1. $x = y$, or
  2. $x$ is existential and there exists $z \in V$ with $(x, z) \in E$ and $\text{apath}(z, y)$, or
  3. $x$ is universal and for all $z \in V$ with $(x, z) \in E$, $\text{apath}(z, y)$ holds.

Status

$\P$-complete (Chandra, Kozen, and Stockmeyer [1], Immerman [2][3]).

Remarks

The reduction to show hardness is from AM2CVP. The original proof simulated an alternating Turing machine (ATM) directly to show that AGAP was complete for ATM logarithmic space [2]. Since $\mathsf{ALOG} = \P$ [1], this showed AGAP was $\P$-complete too. When this problem is generalized to hierarchical graphs it remains in $\P$, provided the graph is “breadth-first ordered”; see Lengauer and Wagner [4]. The proof sketched above also shows that the problem remains $\P$-complete when the partition $(A, B)$ induces a bipartite graph.

When restricted to only existential vertices, the problem is equivalent to the Directed Graph Accessibility Problem, variously called “GAP” and “STCON,” and shown by Savitch to be complete for $\mathsf{NLOG}$ [5]. Peterson shows that the undirected version of AGAP is also $\P$-complete. When restricted to undirected graphs with only existential vertices, this problem is equivalent to the Undirected Graph Accessibility Problem, called “UGAP” or “USTCON,” which is known to be complete for the special case of nondeterministic logarithmic space known as symmetric logarithmic space or $\mathsf{SLOG}$ (Lewis and Papadimitriou [6]).

Yasuura shows a generalization of AGAP is $\P$-complete via a reduction to Path Systems (PATH) [7]. He considers the graph reachability problem on directed hypergraphs, graphs whose edges consist of a subset of vertices and a single vertex.

References

  1. A. K. Chandra, D. C. Kozen, and L. J. Stockmeyer (1981). Alternation.
  2. N. Immerman (1981). Number of quantifiers is better than number of tape cells.
  3. N. Immerman (1987). Languages that capture complexity classes.
  4. T. Lengauer and K. W. Wagner (1988). The binary network flow problem is logspace complete for P.
  5. W. J. Savitch (1970). Relationships between nondeterministic and deterministic tape complexities.
  6. H. R. Lewis and C. H. Papadimitriou (1982). Symmetric space-bounded computation.
  7. H. Yasuura (1983). The reachability problem on directed hypergraphs and computational complexity.
  8. Greenlaw, R., Hoover, H. J., and Ruzzo, W. L. (1995). Limits to Parallel Computation: P-Completeness Theory. Oxford University Press. Problem A.2.3.