Core Concepts
Preprocessing algorithms can identify vertices that belong to an optimal feedback vertex set by finding antler decompositions in the input graph.
Abstract
The paper introduces a new graph structure called an antler decomposition, which can be used to identify vertices that belong to an optimal feedback vertex set in a graph.
The key insights are:
An antler decomposition consists of two disjoint vertex sets (head, antler) such that:
The subgraph induced by head and antler contains a set of |head| vertex-disjoint cycles, certifying that any feedback vertex set must contain at least |head| vertices from head ∪ antler.
The subgraph induced by antler is acyclic, with each tree connected to the rest of the graph by at most one edge.
The authors show that finding a non-empty antler decomposition is NP-hard, but provide an FPT algorithm to find an antler decomposition parameterized by the size of the head.
They generalize the concept to z-antlers, where the subgraph induced by head and antler contains a head-certificate of order z, meaning each component has a feedback vertex set of size at most z.
The authors provide an FPT algorithm to find a z-antler decomposition parameterized by the size of the head and the order z.
These antler decompositions can be used as a preprocessing step to reduce the search space for exact algorithms solving the Feedback Vertex Set problem, by identifying vertices that must be included in an optimal solution.
Stats
fvs(G) = |C| + fvs(G - (C ∪ F))
fvs(G) ≥ |C|
Quotes
"If (C, F) is an antler in G, then fvs(G) = |C| + fvs(G - (C ∪ F))."
"If (C, F) is a z-antler in G for some z ≥ 0, then for any X ⊆ C the pair (C \ X, F) is a z-antler in G - X."