1: \begin{abstract}
2: We give a combinatorial condition for the existence of efficient, LP-based FPT algorithms
3: for a broad class of graph-theoretical optimisation problems.
4: Our condition is based on the notion of biased graphs known from matroid theory.
5: Specifically, we show that given a biased graph $\Psi=(G,\cB)$, where $\cB$ is a class
6: of balanced cycles in $G$, the problem of finding a set $X$ of at most $k$
7: vertices in $G$ which intersects every unbalanced cycle in $G$
8: admits an FPT algorithm using an LP-branching approach, similar to
9: those previously seen for VCSP problems (Wahlstr\"om, SODA 2014).
10: Our algorithm has two parts. First we define a \emph{local problem},
11: where we are additionally given a root vertex $v_0 \in V$ and asked
12: only to delete vertices $X$ (excluding $v_0$) so that the connected component of $v_0$ in $G-X$
13: contains no unbalanced cycle. We show that this local problem admits
14: a persistent, half-integral LP-relaxation
15: with a polynomial-time solvable separation oracle,
16: and can therefore be solved in FPT time via LP-branching,
17: assuming only oracle membership queries for the class of balanced cycles in $G$.
18: We then show that solutions to this local problem
19: can be used to tile the graph, producing an optimal solution to the
20: original, global problem as well.
21: This framework captures many of the problems previously solved
22: via the VCSP approach to LP-branching, as well as new generalisations,
23: such as Group Feedback Vertex Set for infinite groups
24: (e.g., for graphs whose edges are labelled by matrices).
25: A major advantage compared to previous work is that
26: it is immediate to check the applicability of the result for a given problem,
27: whereas
28: testing applicability of the VCSP approach
29: for a specific VCSP,
30: requires determining the existence of
31: an embedding language
32: with certain algebraically defined properties,
33: which is not known to be decidable in general.
34: \end{abstract}
35: