DAG algorithm
shortest path s-> all other nodes
× negative cycles (no shortest path)
- bellmanford
with negative weights
proof: 图中路径 至多 有|v|-1 条边,否则有循环。在无负回路中,可不走此循环以减少距离。
通过对点 出边 的搜索,不断更新最短距离,直到所有点的出路都遍历过。
- dijkstra
cannot work with negative weights
greedy
不断更新已遍历的点集合visited:{}。
topological sorting:
- is not unique
- build system, apt-get tool, which task proceed which one
- stack 放已经访问过的顶点。一顶点的出边全部遍历即为已访问。直到所有顶点访问结束。初始的unvisited点为随机选择。
implement in scala:
import scala.io._ class WeightedDigraph(val n: Integer) { case class Edge(from: Int, to: Int, weight: Long) val adj = Array.fill(n)(List.empty[Edge]) var m = 0 def addEdge(from: Int, to: Int, w: Long) = { adj(from) = Edge(from, to, w)::adj(from) m+=1 this } // Using Dijkstra's algorithm def shortestDistances(s: Int) = { val heap = IndexedHeap[Long](n) def put(v: Int, w: Long) { heap.put(v+1, w) } def deleteMin = heap.deleteMin match { case IndexKeyPair(v, w) => IndexKeyPair(v-1, w) } def changeKey(v: Int, w: Long) { heap.changeKey(v+1, w) } def contains(v: Int) = heap.contains(v+1) def get(v: Int) = heap.get(v+1) val distances = Array.fill(n)(Long.MaxValue) distances(s) = 0 (0 until n) foreach { v => put(v, if (v==s) 0 else Long.MaxValue)} while(!heap.isEmpty) { val pair = deleteMin val v = pair.index val weight = pair.key distances(v) = weight adj(v) filter { edge => contains(edge.to) } foreach { edge => val newDist = distances(v) + edge.weight if (get(edge.to) > newDist) changeKey(edge.to, newDist) } } distances } def floydWarshal: Either[Int, Array[Array[Long]]] = { var prev = Array.ofDim[Long](n, n) val current = Array.ofDim[Long](n, n) // Base case, budget is zero for (i <- 0 until n; j <- 0 until n) { prev(i)(j) = if (i == j) 0l else { adj(i).find(_.to == j).map(_.weight) getOrElse Long.MaxValue } } for (k <- 0 until n) { for (i <- 0 until n; j <- 0 until n) { val inherited = prev(i)(j) val part1 = prev(i)(k) val part2 = prev(k)(j) val composite = if(part1 == Long.MaxValue || part2 == Long.MaxValue) Long.MaxValue else part1 + part2 current(i)(j) = math.min(inherited, composite) if (i == j && current(i)(j) < 0) return Left(i) } prev = current } Right(current) } def reversed: WeightedDigraph = { val r = WeightedDigraph(n) for(u <- 0 until n; edge <- adj(u)) { r.addEdge(edge.to, u, edge.weight) } r } def bellmanFord(start: Int) : Either[Int, Seq[Long]] = { var prev = Array.ofDim[Long](n) val current = Array.ofDim[Long](n) // Base case, length is zero for(v <- 0 until n) { prev(v) = if(v == start) 0l else Long.MaxValue } val r = reversed for(length <- 1 until n; v <- 0 until n) { val inherited = prev(v) val cd = r.adj(v).map { edge => if (prev(edge.to) == Long.MaxValue) Long.MaxValue else prev(edge.to) + edge.weight }.min current(v) = math.min(inherited, cd) prev = current } // Checking for negative cycles, if a path of length N is better then there's a negative cycle for(v <- 0 until n) { val inherited = prev(v) val cd = r.adj(v).map { edge => if (prev(edge.to) == Long.MaxValue) Long.MaxValue else prev(edge.to) + edge.weight }.min if(math.min(inherited, cd) < prev(v)) return Left(v) } Right(current.toSeq) } } object WeightedDigraph { def apply(n: Int) = new WeightedDigraph(n) def fromFile(filename: String) = { val lines = Source.fromFile(filename).getLines() val n = lines.next().split("\\s+")(0).toInt val wg = new WeightedDigraph(n) lines foreach { line => line split "\\s+" match { case Array(u,v,w) => wg.addEdge(u.toInt - 1, v.toInt - 1, w.toInt) } } wg } }
From satrys,
Kristen wang
浙公网安备 33010602011771号