A chordal network is a data structure that represents polynomial ideals in terms of the paths of a certain directed graph. Remarkably, several polynomial ideals with exponentially many components admit compact chordal network representations. Moreover, chordal networks can be efficiently post-processed to compute several properties of the underlying variety, such as cardinality, dimension, components, elimination ideals, and radical ideal membership.
We now present some examples.
Ideal of adjacent minors: Consider the ideal of adjacent minors of a $2\times n$ matrix . This ideal decomposes into $F_n$ components, where $F_n$ is the Fibonacci number. These (exponentially many) components can be represented compactly with a chordal network.
|
|
|
|
|
Once we have the chordal network, one may verify that the variety has codimension 9, and that it has $F_{10}=55$ components.
|
Edge ideals: The edge ideal of a graph $G=(V,E)$ is generated by the monomials $x_ix_j$ for $ij\in E$. Edge ideals have a very nice combinatorial structure, but they often have an exponential number of components. Chordal networks might be used to efficiently describe these large decompositions. The following code computes a chordal network representation for edge ideal of the product graph $C_3\times P_n$.
|
|
|
|
|
|
This variety has codimension 10, and has $48=3\times 2^{5-1}$ components.
|
Chromatic ideal of a cycle: Coloring graphs is a classical NP-hard problem, but it is tractable for certain families of graphs. In particular, coloring the cycle graph $C_n$ is trivial. However, solving the problem algebraically (see chromaticIdeal) can be quite challenging using Gr\"obner bases. On the other hand, this chromatic ideal has a chordal network representation with less than $3n$ nodes [CP2017]. This network can be found with the command chordalTria(N), but the calculation requires Maple (see TriangularDecompAlgorithm).