Cut-sufficiency in series-parallel graphs
The main result in the same paper by Chakrabarty, Fleischer and Weibel [1] characterizes the pairs , with a series-parallel graph, for which there is an integral routing for any satisfying the cut condition. Let be the complete bipartite graph with parts of cardinals and . We call odd-spindle a pair with with even, and is composed of a simple cycle of length spanning the vertices of degree of , plus an edge between the two vertices of degree . For , it gives:
A minor of a pair is a pair obtained from by a sequence of contraction of edges of and deletion of edges of and .
A graph is series-parallel ( is a SP graph) if it can be obtained from an arbitrary tree by a sequence of series and parallel operation. A series operation replace an edge by a path of arbitrarily length, with the same endpoints and all the inner vertices are fresh vertices. A parallel operation replicates an edge in an arbitrary number of edges with the same endpoints. Series-parallel graphs coincides with graphs with no minor , and with graphs of tree-width .
A pair is cut-sufficient if for any such that any instance that satisfies the cut condition admits a (fractional) solution.
Theorem 1: (Chakrabarty , Fleischer and Weibel [1]) If is series-parallel, then is cut-sufficient if and only if has no odd-spindle minors. If is Eulerian and satifies the cut condition, then there is an integral solution.
The proof relies on the congestion-distortion equivalence theorem, and a particular case of cut-sufficiency for series-parallel that was proved in an earlier paper by Chekuri, Shepherd and Weibel [2]. One implication is clear: an odd-spindle with and uniformly equal to is cut-sufficient, though there is no solution. As being cut-sufficient is obviously closed under the minor operations, the cut-sufficiency implies that there is no odd-spindle minor.
Special case :
Let be an instance that satifies the cut condition. We may assume that:
is connected, and even is -connected.
there are no two parallel edges, one in and one in (decrease and on both).
every edge in or is in a tight cut (decrease on or increase on ).
Series-parallel graphs have very strong structural properties. As mentionned above, they are graphs of treewidth 2 , which implies that there are many 2-vertex-cuts. Let's take an arbitrary 2-vertex-cut
, and consider the graph obtained by contracting all the edges of : this graph is a . We start by proving the following. A -instance is an instance
where is a with possibly one additional edge between the two vertices of degree .
Lemma 2: (Chekuri , Shepherd, Weibel [2]) if is a -instance and does not have an odd-spindle as a minor, then is cut-sufficient.
Proof: Recall that denotes the surplus of the cut defined by .
Let satisfy the cut condition. We denote by the two vertices of degree , and the vertices of degree in . By assumption . We may also assume that is tight for any , otherwise we increase the capacity on the edge and decrease it on
. Then simplify the instance by removing parallel edges in and .
Lemma 3: If is bipartite, then is cut-sufficient.
Proof: is bipartite, with parts . Let be a demand with , . We would like to do one of the following operation without violating the cut-condition:
If the first one can't be done, there is a central tight cut containing but not and . If the second one can't be done there is a central tight cut containing and but not
.
We partition into parts to as in the next figure, where is the red set, the blue set. For instance , , and , .
Let and . Then, by counting (and remembering that is a bipartition for and is independant in ):
Because and (cut condition):
The last inequality comes from the existence of the demand . This is a contradiction, hence one of the two operations can be carried out. Also, these operations preserves the integrality of the surplus function, hence at least of the demand can be moved this way at each step. By induction upon the total demand, we reduce the instance to an instance with demand only between , where the cut condition is sufficient.
Suppose . Then, as is tight for any , the instance is not feasible, hence is not bipartite. We get an odd cycle in and so an odd-spindle minor in .
Suppose now that . Then we try to choose for any demand edge , how much of is routed along the path , with respect to using exactly all of the capacities of . This is a capacitated fractional -matching problem on vertices and edges with , and where is an dummy vertex with and is adjacent to every node (including itself) in with infinite capacity. If a solution exists then we easily deduce a solution to the routing problem. Otherwise we know that there exists three sets , and partitioning , with:
Because each vertex in defines a tight set we also have:
Subtracting it from the inequality, we get:
Consider the instance obtained from by adding a demand edge with . If there is a cut with negative excess in , then is a cut with negative excess in , contradiction. But the instance is not feasible: is tight for each so the demand cannot be routed through . Hence by induction contains an odd-spindle as a minor, from which we get that contains an odd-spindle too, using a demand edge in .
Properties of series-parallel graphs
We recall some properties of series-parallel graphs. Remember that a SP graph is a graph with no minor .
Lemma 4: In a SP graph, a simple cycle does not intersect any central cut more than twice.
Proof: By contradiction.
Use the centrality of the cut to get the red paths, this gives a minor.
Lemma 5: In a SP graph, if there are three node-disjoint -paths, then is a -cut.
Proof: Any path between the inner nodes of two distinct paths would give a .
Lemma 6: Let , be a pair of nodes, and be a -cut separating from , , two central sets with and . Then , are central.
Proof:
is connected as and are connected and intersect in . It remains to prove that is connected. By contradiction.
First case : suppose is disconnected from , with in the component of in . Then take the red path by centrality of , the blue path by centrality of , the green path by centrality of (to get to ) and (then to or ). give a .
Second case : suppose is disconnected from , with not in the component of or in . Then again take the red and blue path, then the green to , then use the centrality of and to complete the green path to somewhere (in ) between and and somewhere (in ) between and . this gives a .
Lemma 7: For any -cut in a SP graph and any component of , there is a vertex (not necessarily unique) such that is disconnected.
Proof: By contradiction. We immediately get a .
Lemma 8: Given any -cut in a SP graph , there is a unique acyclic orientation of with being the unique source and the unique sink.
Proof: By recursion, using the previous lemma.
Lemma 9: Orient a SP graph with respect to a -cut . Define the relation if there is a directed path from to
. Then defines a lattice.
Proof: By contradiction. Let be a maximal pair such that there are
and with but and are not comparable. Then there is a .
Lemma 10: Given the acyclic orientation induced by a -cut in a SP graph , any simple cycle can be decomposed into two directed -paths. Then is a -cut.
Proof: By contradiction. Take four maximal consecutive directive paths on a simple cycle . Get as and as . If
we get a , hence . Repeat on all the possible way to choose the four blue paths, we eventually get that
, this contradicts the simplicity of .
We prove the second statement. if it is a -cut. Else there are directed paths from to , from to and from to . Together they define a third disjoint -path. Hence is a -cut.
Special case: compliant instances
We say that an edge is fully compliant if is series-parallel, and that an instance is fully compliant if every demand edge is compliant. The proof of the cut-sufficiency condition for fully compliant instances relies on a -cut reduction that we now describe. Suppose there is a -cut , and a partition
, of the components of , such that
. Then we can reduce the problem into two instances over and respectively. We may have to add a demand edge in one of them, and a supply edge in the other, in order to make the two instances satisfy the cut condition, but there is no difficulty in doing so and proving the correctness of this reduction.
Now it is easy to prove that fully compliant instances are cut-sufficient. We may assume that there is a 2-cut with three disjoint -paths, otherwise is a simple cycle for which cut-sufficiency is well-known (for example by Okamura-Seymour's theorem). As is a vertex cut there is a partition of
, with , and we can assume not all the three paths are in one side. Now, , otherwise we could find a path connecting two of our three
-paths, and this would give a minor. Hence we can apply the -cut reduction. This gives:
Lemma 11: (Chekuri , Shepherd, Weibel [2]) Fully compliant instances are cut-sufficient.
Choose any -cut and consider the associated acyclic orientation. Say that an edge is compliant (for this orientation) if there is a directed -path containing and
. An instance is compliant if there is an acyclic orientation for some -cut such that every demand edge is compliant or fully compliant (it may happen that a fully compliant edge is not compliant, if there are no three node-disjoint
-paths). The next step is:
Lemma 12: (Chekuri , Shepherd, Weibel [2]) Compliant instances are cut-sufficient.
Proof: If all the demand edges are fully compliant we are done. Hence we can assume that there is a demand edge that is not fully compliant, but only compliant. Our goal is to replace by two demand edges such that the new instance is still compliant, satisfies the cut condition and is somehow simpler.
Lemma 13: If an edge is not fully compliant, then there is a -cut separating from , and three disjoint paths from to .
Proof:
has a minor, containing the edge , hence there are vertices and paths joining every possible pairs, one of them containing , say the -path . Then by removing , we get three disjoint -paths. Hence is the -cut that we are looking for.
Because is compliant there is a directed -path going through in that order. Because is not fully compliant there is two cut separating from , and three disjoint -paths. We may choose one of them to contain an path . We may assume that is on between and . must be in the same component as and in , because one of the three -paths does not contain or (and is a -cut). separates from , hence is on , between and or between and . Let's say goes through in that order.
We want to replace the demand edge by two demand edges or . Suppose this is not possible, and we will derive a contradiction. Then we get two central tight cuts and where:
contains but not ,
contains but not .
and are central, so are and by Lemma 6. Consider and its components , and are in two distinct components (as separates them). Suppose some component has two vertices with and V \ setminus . and are connected, hence there are two paths from , one to , one to . and are also connected, hence there are two paths from to and from to . These four paths can be chosen node-disjoint. Moreover, there is an -path in and a -path avoiding . This gives a . Hence each is completely contained in either or .
As the cycle intersects and exactly twice each, we have:
Suppose first that is in the same component as . Then take any demand edge , and consider a directed -path containing and (because is compliant). If and are in two distinct components, and distinct from the component , then intersects at least 5 times and , contradiction. Any demand between two components has one endpoint in .
Suppose now that . Take a component of , and suppose that and are both non-empty, then there is and as evidence. is in some path starting from and is in some path ending at (by the acyclic orientation of ). Because and can be crossed at most twice by a cycle, the situation looks like the following picture:
and are vertex-disjoint: otherwise we could make a directed path from to itself through . This path could then be extended into an -path crossing four times, contradiction. is connected hence there is a path between a node of and a node of , in . Using another component (there are at least two components, and each of them is connected to and by centrality of these sets), and the connectedness of and , we get a .
From this we conclude that every component has (out-component) or (in-component). Moreover any demand edge between two components is contained in an -path which crosses and exactly twice, hence is in an out-component and is in an in-component.
Hence, in both cases, the instance obtained by contracting each component into a single node, as well as and is a with bipartite where is the set of vertices of degree . Moreover the new instance satisfies the cut condition (we only did minor operations), but the cut induced by and in the contracted graph are tight and show that no routing may exists. This contradicts the simultaneous existence of and .
Thus, the demand can be pushed to either or . Then the number of compliant but not fully compliant demand edges is decreased by one. By induction the instance is routable.
General case: odd-spindle free instances
We now give a proof of the general theorem 1 for series-parallel graphs: is cut-sufficient if and only if it does not contain an odd-spindle minor. We only proved on direction, let's prove the other one: we assume that is not cut-sufficient (that is has congestion greater than ) and prove that it contains an odd-spindle.
We start with a consequence of the congestion-distortion equivalence, that can be proved by carefully applying the complementarity slackness conditions in the proof of the congestion-distortion equivalence. Given capacities , demands and a demand edge , we say that a set is a bubble for is is tight and . We say that the set of -paths is covered by bubbles if every path in it intersects at least one bubble for . Then we use the following fact (that we don't prove):
Lemma 14: If is not sufficient, then there are and such that for any demand edge , if does not cross any bubble, then crosses some tight cut more than once. ( and can actually be taken to be optimal solutions for the maximization of congestion problem).
We take and as in the lemma. Now the proof is in two parts: finding a demand with covered by bubbles, and from there finding an odd-spindle.
Lemma 15: There is a demand edge such that is covered by bubbles.
Proof: Orient with respect to some -cut . is not cut-sufficient, hence there is a non-compliant demand . We choose such that is minimal.
By contradiction. Suppose is not covered by bubbles, there is a path that does not cross any bubble, and a central tight cut such that (more than would give a ). Define , , and as in the following picture:
and are chosen such that the two black paths are minimal. Then there are three-disjoint (,)-paths, is a -cut. Moreover, by carefully choosing and , we can ensure that any -path inside crosses some tight cut twice more that (a shortcut by a violating -paths would decrease the number of tight cuts crosses by ).
is the component of containing , and the component containing . By symmetry we may assume and are not in . We want to show the existence of a -cut distinct from , and two disjoint -paths, not containing or , with a demand between their inner nodes. We distinguish two cases:
Suppose there is a tight cut defined by some set , , that intersects every -path in , but not between and . Take inclusion-wise minimal. being a -cut, does not intersect at all. Let be the component of containing , and . cannot be a bubble covering , hence . Then,
From that we get and then . Therefore there is a demand with and . Then we prove that there is a path from to inside : otherwise there would be a component of that does not intersect . Therefore would also cut all the -paths inside , and
Hence , contradicting the minimality of . Using also the centrality of , we easily get to the following situation ( may be equal to ):
must be a -cut because of the three disjoint -paths, and must be in distinct components. Thus there must be an -path containing (in the component of ), disjoint from the -paths containing .
Otherwise there is no single tight cut intersecting all the -paths inside . Choose a minimal family of sets defining central tight cuts that covers all such paths, without intersecting between and . Then all these cuts are bubbles for . Moreover, no two of them are crossed: this would give a , by using . Also the union of them is not tight: this would give a single cut covering the -paths inside . Hence there must be a demand between two of these sets. Take a and a paths, using between and it is then easy to find with three disjoint -paths, one through each of .
Then in both cases is not compliant, and (supposing wlog that ), contradicting the choice of .
We can now conclude the proof of theorem 1 by exhibiting an odd-spindle minor.
Proof: Consider a demand not covered by bubble. Let be a minimal family of bubbles covering every -paths. By centrality of the bubbles, . We distinguish two cases:
. We can assume that the bubbles are disjoint: associate to each bubble a path covered only by that bubble. If two bubbles are crossed, then we can find a path between their associated paths, and using the path associated to a third bubble we get a . Similarly there is no edge in between two bubbles.
Now contract each bubble into a single node, as well as the components of and in . It is easy to check this operation gives a , with each vertex of degree defining a tight cut, and a positive demand between the vertices of degree . This instance also satisfies the cut condition but does not have a routing, hence there must be an odd-spindle minor.
, . Let be the component of containing , and the component containing . Then:
If , the RHS becomes , contradiction. Hence contains at least one vertex . By centrality of and we can find two -paths: disjoint from , and disjoint from , and a (minimal) path from to inside and a (minimal) path from to inside . We deduce that there are three disjoint -paths, is a -cut. Applying Lemma 6, and are both central.
Contract every edge not in , this gives a that satisfies the cut condition. But the cuts induced by and shows that this instance is not feasible: there is an odd-spindle again.
This concludes the proof of Theorem 1.
1. | Amit Chakrabarti, Lisa Fleischer, Christophe Weibel: When the cut condition is enough: a complete characterization for multiflow problems in series-parallel networks. In: Proceedings of the 44th symposium on Theory of Computing. , 19 -- 26. (2012). | |
2. | Chandra Chekuri, F. Bruce Shepherd, Christophe Weibel: Flow-cut gaps for integer and fractional multiflows. Journal of Combinatorial Theory, Series B. Elsevier, (2012). | |