Similarly, an induced cycle is a cycle that is an induced subgraph of G; induced cycles are also called chordless cycles or (when the length of the cycle is four or more) holes. An antihole is a hole in the complement of G, i.e., an antihole is a complement of a hole.
The length of the longest induced path in a graph has sometimes been called the detour number of the graph. The induced path number of a graph G is the smallest number of induced paths into which the vertices of the graph may be partitioned, and the closely related path cover number of G is the smallest number of induced paths that together include all vertices of G. The girth of a graph is the length of its shortest cycle, but this cycle must be an induced cycle as any chord could be used to produce a shorter cycle; for similar reasons the odd girth of a graph is also the length of its shortest odd induced cycle.
The illustration shows a cube, a graph with eight vertices and twelve edges, and an induced path of length four in this graph. A straightforward case analysis shows that there can be no longer induced path in the cube, although it has an induced cycle of length six. The problem of finding the longest induced path or cycle in a hypercube, first posed by , is known as the snake-in-the-box problem, and it has been studied extensively due to its applications in coding theory and engineering.
Many important graph families can be characterized in terms of the induced paths or cycles of the graphs in the family.
It is NP-complete to determine, for a graph G and parameter k, whether the graph has an induced path of length at least k. credit this result to an unpublished communication of Yannakakis. However, this problem can be solved in polynomial time for certain graph families, such as asteroidal-triple-free graphs or graphs with no long holes.
It is also NP-complete to determine whether the vertices of a graph can be partitioned into two induced paths, or two induced cycles. As a consequence, determining the induced path number of a graph is NP-hard.
The complexity of approximating the longest induced path or cycle problems can be related to that of finding large independent sets in graphs, by the following reduction. From any graph G with n vertices, form another graph H with twice as many vertices as G, by adding to G an independent set I of n vertices, and an edge from each vertex of G to each vertex of I. Then if G has an independent set of size k, H must have an induced path and an induced cycle of length 2k, formed by alternating vertices of the independent set in G with vertices of I. Conversely, if H has an induced path or cycle of length k, any maximal set of nonadjacent vertices in G from this path or cycle forms an independent set in G of size at least k/3. Thus, the size of the maximum independent set in G is within a constant factor of the size of the longest induced path and the longest induced cycle in H. Therefore, by the results of on inapproximability of independent sets, unless NP=ZPP, there does not exist a polynomial time algorithm for approximating the longest induced path or the longest induced cycle to within a factor of O(n1-ε) of the optimal solution.
Holes (and antiholes in graphs without chordless cycles of length 5) in a graph with n vertices and m edges may be detected in time (n+m2)