整数规划中的 Well-Solved Problems
摘要
整数规划问题中有很多Well-Solved Problems,对于这些问题已经有很多研究者提出了非常有效的算法来进行求解。面对新出现的问题,很多时候可以将其分解为子问题,若这些子问题属于某一类Well-Solved Problems,那将起到非常显著的求解效果。
Properties of Easy Problems
考虑一个组合优化问题(Combinatorial optimization problem,以下简称COP)
C
O
P
:
m
a
x
{
c
x
:
x
∈
X
⊆
R
n
}
COP:max\{cx: x \in X \subseteq \mathbb{R}^n\}
COP:max{cx:x∈X⊆Rn}
我们一般会考虑它的 dual problem,同时也可以考虑其相关的 separation problem.
定义1 与COP相关的 separation problem:给定 x ∗ ∈ R n x^* \in \mathbb{R}^n x∗∈Rn,是否满足 x ∗ ∈ c o n v ( X ) ? x^* \in conv(X)? x∗∈conv(X)? 若不满足,则找到不等式 π x ≤ π 0 \pi x \leq \pi_{0} πx≤π0,该不等式被所有 X X X中的点满足,但 x ∗ x^* x∗违反该不等式.
考虑一个问题是否存在高效算法,以下四个属性基本上是同时存在的:
- Efficient Optimization Property: For a given class of optimization problems ( P P P) m a x { c x ∶ x ∈ X ⊆ R n } max\{cx ∶ x ∈ X ⊆ ℝ^n\} max{cx∶x∈X⊆Rn}, there exists an efficient (polynomial) algorithm.
- Strong Dual Property: For the given problem class, there exists a strong dual problem
(
D
)
(D)
(D)
m
i
n
{
ω
(
u
)
:
u
∈
U
}
min\{\omega(u): u \in U\}
min{ω(u):u∈U} allowing us to obtain optimality conditions that can be quickly verified:
x ∗ ∈ X x^∗ ∈ X x∗∈X is optimal in P P P if and only if there exists u ∗ ∈ U u^∗ ∈ U u∗∈Uwith c x ∗ = ω ( u ∗ ) cx^∗ = \omega(u^∗) cx∗=ω(u∗). - Efficient Separation Property: There exists an efficient algorithm for the separation problem associated with the problem class.
- Explicit Convex Hull Property: A compact description of the convex hull c o n v ( X ) conv(X) conv(X) is known, which in principle allows us to replace every instance by the linear program: m a x { c x : x ∈ c o n v ( X ) } max\{cx : x ∈ conv(X)\} max{cx:x∈conv(X)}.
Wolsey 在书上表示存4⇒2⇒3的关系,但我还没完全搞清楚,基础部分还要再看看
几类Well-Solved Problems
1.具有全单模矩阵的IP
2. Minimum Cost Network Flows
3. Special Minimum Cost Flows
3.1 Shortest Path
3.2 Maximum s − t Flow
4. Optimal Trees
5. Submodularity and Matroids
6. Two Harder Network Flow Problems
6.1 The Steiner Tree Problem
6.2 The Fixed Charge Network Flow Problem
总结
目前课题暂时没用上,待续