Algorithm for graph homomorphisms
Status partial medium confidence
The original question of whether there is an algorithm deciding the existence of a homomorphism from $G$ to $H$ in time $O(c^{|V(G)|+|V(H)|})$ for some absolute constant $c$ remains open in full generality. Since 2010, several works have extended the list of graph classes admitting such plain-exponential algorithms (e.g., Wahlstrom-type results were generalized by Dadsetan and Bulatov to further classes for the counting version, and Rzazewski gave an $O^{*}((b+2)^{|V(G)|})$ algorithm parametrized by the bandwidth of the complement of $H$), and Fomin, Golovnev, Kulikov, and Mihajlin proved an ETH lower bound of $2^{\Omega(n \log h / \log\log h)}$ for the general problem, which however does not rule out a $c^{n+h}$ algorithm.
Cited literature (4)
-
Proves an ETH-based lower bound of $2^{\Omega(n \log h / \log\log h)}$ for deciding graph homomorphism, ruling out single-exponential-in-$n\log h$ algorithms but not a $c^{|V(G)|+|V(H)|}$ algorithm.
-
Gives an exact algorithm for graph homomorphism running in time $O^{*}((b+2)^{|V(G)|})$, where $b$ is the bandwidth of the complement of $H$, expanding the list of classes for which the problem is solvable in plain-exponential time.
-
Generalizes Wahlstrom's result for counting graph homomorphisms in time $k^{|V(G)|+|V(H)|}\cdot\mathrm{poly}$ from bounded clique-width to additional classes of target graphs $H$, enlarging the known cases of plain-exponential algorithms.
-
Provides a fine-grained classification of the homomorphism problem in time $O^{*}(s(H)^{cw})$ with matching SETH-based lower bounds, refining the picture for restricted target graphs $H$ (clique-width-parametrized) but leaving the general $c^{|V(G)|+|V(H)|}$ question open.
Reviewer notes. The original question concerns the decision version with constant base $c$ independent of $G,H$. The 2015 ETH lower bound of $2^{\Omega(n \log h / \log\log h)}$ does NOT rule out a $c^{|V(G)|+|V(H)|}$ algorithm (e.g., when $h \gg n$, $c^{n+h}$ can exceed the lower bound). All cited papers extend the known classes of $H$ for which plain-exponential algorithms exist, or give lower bounds for restricted parameterizations, but none affirmatively or negatively resolves the original FHK question. The exact published year/venue of arXiv:1810.03087 was not verified beyond the preprint.
Discussion
An affirmative answer is known in several cases: if $ H=K_k $ (graph coloring) [L], [BH], [K]; if $ H $ has bounded treewidth [FHK]; if $ H $ has bounded cliquewidth [W].
Bibliography
-
[BH]Andreas Björklund, Thore Husfeldt: Inclusion--Exclusion Algorithms for Counting Set Partitions, Proc. FOCS'06 (2006). -
★
[FHK]Fedor V. Fomin, Pinar Heggernes, Dieter Kratsch: Exact Algorithms for Graph Homomorphisms, Theory Comput. Syst. 41 (2007), no. 2, 381--393. MathSciNet MathSciNet -
[K]Mikko Koivisto: An $ O^\ast(2^n) $ Algorithm for Graph Coloring and Other Partitioning Problems via Inclusion--Exclusion, Proc. FOCS'06 (2006). -
[L]Eugene L. Lawler: A note on the complexity of the chromatic number problem, Information Processing Lett. 5 (1976), no. 3, 66--67. MathSciNet MathSciNet -
[W]Magnus Wahlström: New Plain-Exponential Time Classs for Graph Homomorphism, CSR2009, LNCS5675 (2009), 346--355.