Companion rule for the model proposal in #1046.
Source / Target
Motivation
This is the canonical structural reduction from hybridization number to directed feedback vertex set ("Cycle Killer", van Iersel–Kelk–Lekić–Scornavacca 2012). It
- connects
MaximumAcyclicAgreementForest to the existing MinimumFeedbackVertexSet hub and transitively to the ILP/QUBO clusters;
- preserves the optimum exactly: the hybridization number of $(T_1, T_2)$ equals the minimum DFVS size of the constructed auxiliary digraph; and
- establishes a 2-approximation bridge (DFVS is 2-approximable on the auxiliary graph, yielding a 2-approximation for hybridization).
Reference
Van Iersel, Kelk, Lekić, Scornavacca, "Cycle Killer…Qu'est-ce que c'est? On the Comparative Approximability of Hybridization Number and Directed Feedback Vertex Set", SIAM Journal on Discrete Mathematics 26(4), 2012. https://epubs.siam.org/doi/10.1137/120864350
Reduction Algorithm
Let $T_1, T_2$ be the input rooted binary phylogenetic trees on leaf set $X$, $n = |X|$.
-
Kernelization (optional but standard). Apply the subtree and chain reductions of Bordewich–Semple (2005) to $(T_1, T_2)$ exhaustively: collapse every maximal common pendant subtree to a single leaf, and replace every maximal common chain of length $\geq 3$ by a chain of length $3$. These reductions preserve hybridization number.
-
Build the display graph $D(T_1, T_2)$: the vertex-disjoint union of $T_1$ and $T_2$ with corresponding leaves identified, plus both roots as sources.
-
Construct the auxiliary digraph $H$. Following §4 of van Iersel et al., introduce one vertex of $H$ for each edge-pair $(e_1, e_2) \in E(T_1) \times E(T_2)$, plus one vertex for each leaf $\ell \in X$. Add a directed arc $u \to v$ between auxiliary vertices whenever the corresponding edge-pairs induce an ancestor relation on their common leaf set in $D(T_1, T_2)$.
-
Reduce. Return
MinimumFeedbackVertexSet on $H$ with unit weights $w \equiv 1$.
Solution lifting. A minimum feedback vertex set $S \subseteq V(H)$ determines a minimum set of edge-pairs whose removal makes the ancestor digraph acyclic; the connected components of the cut $D(T_1, T_2)$ form a maximum acyclic agreement forest $F$ with $|F| = |S| + 1$.
Size Overhead
Target size fields from pred show MinimumFeedbackVertexSet → size_fields = [num_arcs, num_vertices].
| Field |
Formula |
num_vertices |
num_leaves ^ 2 |
num_arcs |
num_leaves ^ 3 |
(After kernelization the effective sizes are bounded by a function of the hybridization number $k$, but the worst-case formulas above express the raw construction.)
Validation Method
Closed-loop test: construct the source instance, apply the reduction, solve the target by brute force, lift the solution, and check that the recovered $F$ is (a) a valid acyclic agreement forest of $(T_1, T_2)$ and (b) $|F| - 1$ matches the ground-truth hybridization number.
Example (fully worked)
Source. $X = {a, b, c, d}$, $T_1 = ((a,b),(c,d))$, $T_2 = ((a,c),(b,d))$ (the 4-leaf quartet swap).
Construction. No pendant subtree or chain is common, so kernelization does nothing. $T_1$ and $T_2$ each have 6 edges (3 internal + 3 leaf-edges after root edges are dropped). The auxiliary digraph $H$ has $6 \cdot 6 + 4 = 40$ vertices and roughly $6^3 = 216$ arcs (upper bound).
Target solved. Minimum DFVS size of $H$ is $|S| = 3$.
Lifted solution. $|F| = |S| + 1 = 4$, and the only acyclic agreement forest achieving this is $F = {{a},{b},{c},{d}}$. Hybridization number = 3.
BibTeX
@article{vanIerselKLS2012,
author = {van Iersel, Leo and Kelk, Steven and Leki\'{c}, Nela and Scornavacca, Celine},
title = {Cycle Killer\ldots Qu'est-ce que c'est? On the Comparative Approximability of Hybridization Number and Directed Feedback Vertex Set},
journal = {SIAM Journal on Discrete Mathematics},
volume = {26},
number = {4},
pages = {1635--1656},
year = {2012},
doi = {10.1137/120864350},
}
@article{BordewichS2005,
author = {Bordewich, Magnus and Semple, Charles},
title = {On the computational complexity of the rooted subtree prune and regraft distance},
journal = {Annals of Combinatorics},
volume = {8},
number = {4},
pages = {409--423},
year = {2005},
doi = {10.1007/s00026-004-0229-z},
}
Companion rule for the model proposal in #1046.
Source / Target
MaximumAcyclicAgreementForest(default variant) — see [Model] MaximumAcyclicAgreementForest #1046MinimumFeedbackVertexSet(directed graph, weighti32)Motivation
This is the canonical structural reduction from hybridization number to directed feedback vertex set ("Cycle Killer", van Iersel–Kelk–Lekić–Scornavacca 2012). It
MaximumAcyclicAgreementForestto the existingMinimumFeedbackVertexSethub and transitively to the ILP/QUBO clusters;Reference
Van Iersel, Kelk, Lekić, Scornavacca, "Cycle Killer…Qu'est-ce que c'est? On the Comparative Approximability of Hybridization Number and Directed Feedback Vertex Set", SIAM Journal on Discrete Mathematics 26(4), 2012. https://epubs.siam.org/doi/10.1137/120864350
Reduction Algorithm
Let$T_1, T_2$ be the input rooted binary phylogenetic trees on leaf set $X$ , $n = |X|$ .
MinimumFeedbackVertexSetonSolution lifting. A minimum feedback vertex set$S \subseteq V(H)$ determines a minimum set of edge-pairs whose removal makes the ancestor digraph acyclic; the connected components of the cut $D(T_1, T_2)$ form a maximum acyclic agreement forest $F$ with $|F| = |S| + 1$ .
Size Overhead
Target size fields from
pred show MinimumFeedbackVertexSet→size_fields = [num_arcs, num_vertices].num_verticesnum_leaves ^ 2num_arcsnum_leaves ^ 3(After kernelization the effective sizes are bounded by a function of the hybridization number$k$ , but the worst-case formulas above express the raw construction.)
Validation Method
Closed-loop test: construct the source instance, apply the reduction, solve the target by brute force, lift the solution, and check that the recovered$F$ is (a) a valid acyclic agreement forest of $(T_1, T_2)$ and (b) $|F| - 1$ matches the ground-truth hybridization number.
Example (fully worked)
Source.$X = {a, b, c, d}$ , $T_1 = ((a,b),(c,d))$ , $T_2 = ((a,c),(b,d))$ (the 4-leaf quartet swap).
Construction. No pendant subtree or chain is common, so kernelization does nothing.$T_1$ and $T_2$ each have 6 edges (3 internal + 3 leaf-edges after root edges are dropped). The auxiliary digraph $H$ has $6 \cdot 6 + 4 = 40$ vertices and roughly $6^3 = 216$ arcs (upper bound).
Target solved. Minimum DFVS size of$H$ is $|S| = 3$ .
Lifted solution.$|F| = |S| + 1 = 4$ , and the only acyclic agreement forest achieving this is $F = {{a},{b},{c},{d}}$ . Hybridization number = 3.
BibTeX