Terminal-set minimum cuts #
A terminal cut is a finite vertex set containing all source terminals and no sink terminals. For a set function with linearly ordered values on a finite vertex type, disjoint terminal sets guarantee attainment of a minimum. For a submodular function, the minimizing sides form a distributive sublattice under inclusion; on a finite vertex type with linearly ordered values, it is finite, with a smallest and a largest side. These extremal sides describe which vertices occur in every minimum cut and which occur in at least one.
For two vertices, pairMinimumValue f s t is the minimum cut value λ(s, t), with λ(s, s) = 0.
It satisfies the ultrametric inequality min (λ(s, v)) (λ(v, t)) ≤ λ(s, t) for distinct s, t.
The theory is independent of a graph representation and does not require symmetry of the set function. Either terminal set may be empty.
An admissible source side contains A and avoids B.
Instances For
The characteristic conditions for a terminal cut.
Every source terminal belongs to an admissible side.
An admissible side contains no sink terminal.
With a finite vertex type, admissible sides form the interval from A to Bᶜ.
Complementing a source side exchanges the source and sink terminal sets.
The complement of an admissible side is admissible for the exchanged terminal sets.
Disjoint terminal sets admit the source terminal set itself as a cut.
Admissibility is closed under union.
Admissibility is closed under intersection.
For disjoint terminal sets, adding A to any vertex set and removing B gives an admissible
side.
Every minimum side is an admissible cut.
The value at a minimum side bounds the value at any admissible cut.
All minimum cuts have the same value.
The attained minimum value among cuts separating the disjoint terminal sets.
Instances For
The minimum value is no larger than the value of any admissible cut.
An admissible side is minimum exactly when its value is the minimum value.
Suppose that g ≤ f and that g is at least min f b everywhere. Then f and g agree on
every set whose f-value is below b.
Suppose that g ≤ f, that g is at least min f b everywhere, and that some admissible side
has f-value below b. Then f and g agree on every minimum side of f.
Suppose that g ≤ f, that g is at least min f b everywhere, and that some admissible side
has f-value below b. Then f and g have the same minimum sides. This is the effect on
minimum cuts of truncating all values above b.
Under the hypotheses of TauCeti.TerminalCut.isMinimum_iff_of_le_of_min_le, the set
functions f and g have the same minimum value.
An injective relabeling preserves the conditions on terminal cuts.
Relabeling all vertices preserves minimum cuts whenever it preserves the set function.
Relabeling vertices does not change the minimum cut value.
The minimum terminal cuts, as a sublattice of finite vertex sets. Its induced lattice is distributive, and is finite when the vertex type is finite.
Instances For
Membership in the cut lattice is precisely minimum-cut optimality.
The union of minimum cuts is minimum.
The intersection of minimum cuts is minimum.
A value-preserving vertex equivalence identifies the lattices of minimum cuts.
Instances For
The minimum-cut order isomorphism maps each source side by the vertex equivalence.
The inverse minimum-cut isomorphism relabels by the inverse vertex equivalence.
The finite nonempty lattice of minimum cuts has bottom and top elements.
Instances For
Minimum terminal cuts have a least side under inclusion.
Minimum terminal cuts have a greatest side under inclusion.
The smallest minimum source side.
Instances For
The largest minimum source side.
Instances For
The smallest side is minimum and contained in every minimum side.
The largest side is minimum and contains every minimum side.
The smallest minimum side belongs to the minimizer lattice.
The largest minimum side belongs to the minimizer lattice.
The least property characterizes the smallest side uniquely.
The greatest property characterizes the largest side uniquely.
The bottom of the bounded minimizer lattice is the smallest minimum source side.
The top of the bounded minimizer lattice is the largest minimum source side.
Relabeling vertices transports the smallest minimum source side.
Relabeling vertices transports the largest minimum source side.
A vertex belongs to the smallest cut exactly when it belongs to every minimum cut.
A vertex belongs to the largest cut exactly when it belongs to some minimum cut.
There is a unique minimum cut exactly when the smallest and largest sides agree.
A side admissible for a pair of singleton terminal sets separates distinct terminals.
The minimum cut value λ(s, t) between two vertices: the minimum value of a cut containing
s and not t when s ≠ t, and zero on the diagonal.
Instances For
The minimum cut value between a vertex and itself is zero by convention.
Between distinct vertices, the minimum cut value is that of the singleton terminal sets.
The minimum cut value between two vertices is no larger than the value of any cut separating them.
A cut separating two vertices is minimum exactly when it attains their minimum cut value.
Ultrametric inequality for minimum cut values: every cut separating s from t
separates s from v or v from t. The intermediate vertex v is arbitrary.
Examples #
On six vertices, charge one unit for including vertex 4, and nothing for the other vertices.
With sources 0, 1 and sinks 2, 3, the two minimum source sides are {0, 1} and {0, 1, 5}.
This nonconstant modular function also illustrates that empty terminal sets cause no exception
to attainment or to the smallest and largest cut formulas.