Shortest Paths in a Graph

The Problem

Let $G=(V,E)$ be a directed graph. Assume that each edge $(i,j)\in E$ has associated weight $c_{ij}$.

Dijkstra’s Algorithm

第一种写法

Dijkstra’s Algorithm: Find shortest paths in graphs with positive edge costs from node $s$.

假设 each edge $e$ has length $l_e\ge 0$. For a path $P$, the length of $P$ - denoted $l(P)$ - is the sum of the length of all edges in $P$.

此外，引进两个集合 $S$ 和 $U$：

• $S$ 的作用是记录已求出最短路径的顶点 $u$, 以及相应的最短路径长度 $d(u)$。
• $U$ 的作用是记录还未求出最短路径的顶点 $v$，那么可以推出

$$d^{\prime }(v)=\underset{e=(u,v):u\in S}{\min }d(u)+l_e$$

---

Dijkstra’s Algorithm $(G,l)$

\quad Let $S$ be the set of explored nodes
\quad \quad For each $u\in S$, we store a distance $d(u)$

\quad Initially S = { s } S=\{s\} S={s} and $d(s)=0$
\quad While $S\ne V$
\quad \quad Select a node $v\notin S$ with at least one edge from $S$ for which $d^{\prime }(v)={\min }_{e=(u,v):u\in S}d(u)+l_e$ is as small as possible
\quad Add $v$ to $S$ and define $d(v)=d^{\prime }(v)$
\quad EndWhile

---

第二种写法

Calculate length of shortest path from the start node $v$ to each node.

---

伪代码：

---

More Complex Problem

Find the shortest paths when costs may be negative.

Negative cost 的实际意义：nodes 代表 agents in a financial setting, $c_{ij}$ 代表 the cost of a transaction in which we buy from agent $i$ and then immediately sell to agent $j$. 这种设定下，a path would represent a succession of transactions, and edges with negative costs would represent transactions that result in profits.

这里讨论两种问题：

1. Given a graph $G$ with weights, decide if $G$ has a negative cycle - that is, a direct cycle $C$ such that
   $$\sum _{ij\in C}{c}_{ij}<0$$
   ($ij$: 由 node $i$ 和 node $j$ 构成的 edge)
2. If the graph has no negative cycles, find a path $P$ from an origin node $s$ to a destination node $t$ with minimum total cost:
   $$\sum _{ij\in P}{c}_{ij}$$

Consider the Minimum-Cost $s$-$t$ Path Problem under the assumption that there are no negative cycles： 因为，假设存在一个 negative cycle、一条从 $s$ 到 cycle 的 path $P_s$、和一条从 cycle 到 $t$ 的 path $P_t$，那么如果无限次绕这个 negative cycle，从 $s$ 到 $t$ 的 total cost 将变得无限小。

Design the Algorithm

如果用 Dijkstra 的思路，每一次将与 $S$ 相邻且到 node $s$ 距离最近的 node $v$ 添加进 $S$，在有 negative weights 的情况下不成立。因为可以走一条开始 cost 贵的 path，但是后部分 path 是 negative cost，最终导致 total cost 小于开始 cost 便宜的 path。
如果每一次计算 cost 的时候都同时加上一个很大的数 $M$，即 $c^{\prime }=c+M$，但是每条 path 所用的 edge 的个数可能不一样，导致额外加的 $M$ 的个数不同。

定理 (6.22)：If $G$ has no negative cycles, then there is a shortest path from $s$ to $t$ that is simple (i.e., does not repeat nodes), and hence has at most $n-1$ edges.

证明：
$\because$ every cycle has nonnegative cost
$\therefore$ the shortest path $P$ from $s$ to $t$ with the fewest number of edges does not repeat any vertex $v$.
若 $P$ 中有重复的 vertex $v$
We could remove the portion of $P$ between consecutive visits to $v$, resulting in a path of no greater cost and fewer edges.
$G$ 有 $n$ 个 nodes。
$\because$ $G$ 是 connected 且 does not contain a cycle
$\therefore$ $G$ has $n-1$ edges

用 $OPT(i,v)$ 表示 the minimum cost of a $v$-$t$ path using at most $i$ edges.

Original Problem: $OPT(n-1,s)$
Fix an optimal path $P$ representing $OPT(i,v)$ as depicted below:

• If the path $P$ uses at most $i-1$ edges, then $OPT(i,v)=OPT(i-1,v)$.
• If the path $P$ uses $i$ edges, and the first edge is $(v,w)$, then $OPT(i,v)=c_{vw}+OPT(i-1,w)$

定理 (6.23): If $i>0$, then
$$OPT(i,v)=\min (OPT(i-1,v),\underset{w\in V}{\min }(OPT(i-1,w)+c_{vw}\left)\right)$$

算法伪代码

---

Shortest-Path($G,s,t$)
\quad $n$ = number of nodes in $G$
\quad Array $M[0...n-1,V]$
\quad Define $M[0...n-1,t]=0$ and $M[0,v]=\infty$ for all other $v\in V$
\quad For $i=1,...,n-1$
\quad \quad For $v\in V$ in any order
\quad \quad \quad Compute $M[i,v]$ using recurrence $(6.23)$
\quad \quad EndFor
\quad EndFor
\quad Return $M[n-1,s]$

---

定理 (6.24): The Shortest-Path method correctly computes the minimum cost of an $s$-$t$ path in any graph that has no negative cycles, and runs in $O(n^3)$ times.

例子

图(a) 所示是一个 directed graph

the Shortest- Path Algorithm constructs the dynamic programming table:

Extensions: Some Basic Improvements to the Algorithm

An Improved Running-Time Analysis

如果有 $n$ 个 nodes，一个 directed graph 可以有大约 $n^2$ 个edges（因为每两个 nodes 之间就可以有一条 edge）。然而一般的 graph 比这种稀疏很多。我们假设有 $m$ 条 edges，那么有

定理 (6.25): The Shortest-Path method can be implemented in $O(mn)$ time.

证明：
对于 Array Entry $M[i,v]$, 由 recurrence $(6.23)$ 可得
$$M[i,v]=\min (M[i-1,v],\underset{w\in V}{\min }(M[i-1,w]+c_{vw}\left)\right)$$ We assumed it could take up to $O(n)$ time to compute this minimum, since there are $n$ possible nodes $w$.
但是这里只用计算和 $n$ 相连的 $w$（因为是 directed graph，只考虑由 $n$ 指向 $w$） 的 minimum.
假设有 $n_v$ denotes the number of edges leaving $v$，那么 it takes time $O(nv)$ to compute the array entry $M[i,v]$.
We have to compute an entry for every node $v$ and every index $0\le i\le n-1$, so this gives a running-time bound of
$$O\left(n\sum _{v\in V}{n}_v\right)=O(nm)$$
$\because$ 是 directed graph，即 ${\sum }_{v\in V}{n}_v$ 中每一条 edge 都只被数了一次
$\therefore$ ${\sum }_{v\in V}{n}_v=m$

Improving the Memory Requirements

Array $M$ has size $n^2$, 这里可以将其 reduce 到 $O(n)$. Rather than recording $M[i,v]$ for each value $i$, we will use and update a single value $M[v]$ for each node $v$, the length of the shortest path from $v$ to $t$ that we have found so far.

依旧保持 $i=1,2,...,n-1$ 的 iteration，但是这里 $i$ 只作为一个 counter。In each iteration, and for each node v, we perform the update
$$M[v]=\min (M[v],\underset{w\in V}{\min }(c_{vw}+M[w]))$$

定理 (6.26): Throughout the algorithm $M[v]$ is the length of some path from $v$ to $t$, and after $i$ rounds of updates the value $M[v]$ is no larger than the length of the shortest path from $v$ to $t$ using at most $i$ edges.

Finding the Shortest Paths

How to recover the shortest path?

• Enhance the code by having each node $v$ maintain the first node after itself ($v$ 之后的第一个 node，记作 $first[v]$) on its path to the destination $t$.
• 当 $M[v]$ is updated，同时 update $first[v]$。
  • Whenever the value of $M[v]$ is reset to the minimum ${\min }_{w\in V}(c_{vw}+M[w])$, we set $first[v]$ to the node $w$ that attains this minimum.

Let $P$ denote the directed “pointer graph” whose nodes are $V$, and whose edges are { ( v , f i r s t [ v ] ) } \{(v, first[v])\} {(v,first[v])}

定理 (6.27): If the pointer graph $P$ contains a cycle, then this cycle must have negative cost.

证明
如果 $first[v]=w$，那么 $M[v]\ge c_{vw}+M[w]$ （因为 $\min$ 中我们取了含有 $w$ 的项）
Let $v_1,v_2,...,v_k$ be the nodes along the cycle $C$ in the pointer graph, and assume that $(v_k,v_1)$ is the last edge to have been added.
for $i=1,2,...,k-1$, $M[v_i]\ge c_{v_i{v}_{i+1}}$
for $i=k$, $M[k]>c_{v_k{v}_1}+M[v_1]$, since we are about to update $M[v_k]$ and change $first[v_k]$ to $v1$.
将 $k$ 个不等式相加
$$0>\sum _{i=1}^{k-1}{c}_{v_i{v}_{i+1}}+c_{v_k{v}_1}$$ a negative cycle

定理 (6.28): Suppose $G$ has no negative cycles, and consider the pointer graph $P$ at the termination of the algorithm. For each node $v$, the path in $P$ from $v$ to $t$ is a shortest $v$-$t$ path in $G$.

证明：
Consider a node $v$ and $first[v]=w$
因为算法停止，所以 $M[v]=c_{vw}+M[w]$
$\because$ $M[t]=0$ [这个是为什么？难道没有指向 $t$ 的 node？]
$\therefore$ the length of the path traced out by the pointer graph is exactly $M[v]$, which we know is the shortest-path distance.