Question

我使用R igraph实现了加权DAG的最长路径计算。

我的实现（如下所示）对于大型图表来说速度很慢。我非常感谢任何加快速度的提示。关于我的实现与最着名/典型算法的距离的任何想法也是受欢迎的。

谢谢！

# g is the igraph DAG
# g <- graph.tree(10000, 2, mode="out")
# E(g)$weight <- round(runif(length(E(g))),2) * 50 
# Topological sort
tsg <- topological.sort(g)    
# Set root path attributes
# Root distance
V(g)[tsg[1]]$rdist <- 0
# Path to root
V(g)[tsg[1]]$rpath <- tsg[1]
# Get longest path from root to every node        
for(node in tsg[-1])
{
  # Get distance from node's predecessors
  w <- E(g)[to(node)]$weight
  # Get distance from root to node's predecessors
  d <- V(g)[nei(node,mode="in")]$rdist
  # Add distances (assuming one-one corr.)
  wd <- w+d
  # Set node's distance from root to max of added distances 
  mwd <- max(wd)
  V(g)[node]$rdist <- mwd
  # Set node's path from root to path of max of added distances
  mwdn <- as.vector(V(g)[nei(node,mode="in")])[match(mwd,wd)]
  V(g)[node]$rpath <- list(c(unlist(V(g)[mwdn]$rpath), node))      
}
# Longest path length is the largest distance from root
lpl <- max(V(g)$rdist)    
# Enumerate longest path
lpm <- unlist(V(g)[match(lpl,V(g)$rdist)]$rpath)    
V(g)$critical <- 0
g <- set.vertex.attribute(g, name="critical", index=lpm, value=1)

Answer 1

我也有一个缓慢的R版本。 200k边缘和30k顶点需要大约20分钟，因此我对具有负边缘权重的图形进行了分解并实现了get.shortest.paths()，您可以通过反转所有边缘权重来查找最长路径。您可以尝试我的R igraph here。

从R实现切换到C时，我经历了100x到1000x的加速。

使用R igraph的加权DAG的最长路径

1 个答案: