r - 找到反向边并从其相反方向减去其权重

Question

我有一个像下面这样的矩阵

m <- expand.grid(LETTERS[1:24],LETTERS[1:24])
m$weight <- runif(nrow(m), 0.01, max = 1)
m <- m[m$Var1!=m$Var2, ] ##remove loop edges
colnames(m) = c("to","from","weight")

并以这种形式描述了一个有向图。我想要做的是减去每对逆边并取其绝对值，并创建一个描述新无向图的新矩阵。即:

abs( edge_weight(A,B) - edge_weight(B,A) )

但我不知道如何只考虑每对一次。

Answer 1

使用igraph

library(igraph)

#dataframe to directed graph
directed_graph <- graph.data.frame(m, directed = T)

#convert to undirected graph by applying desired function
undirected_graph <- as.undirected(directed_graph, mode = "collapse", 
                                  edge.attr.comb = list(weight = function(x) abs(x[1] - x[2])))

#final result
df <- as.data.frame(cbind(get.edgelist(undirected_graph), 
                          unlist(get.edge.attribute(undirected_graph))))
colnames(df) <- c("edge1", "edge2", "weight")
rownames(df) <- NULL

给出

> head(df)
  edge1 edge2             weight
1     B     C  0.310624073725194
2     B     D  0.587582074650563
3     C     D 0.0327853348944336
4     B     E   0.19360910307616
5     C     E  0.328824346032925
6     D     E   0.13037203295622

示例数据:

set.seed(123)

m <- expand.grid(LETTERS[1:24], LETTERS[1:24])
m$weight <- runif(nrow(m), 0.01, max = 1)
m <- m[m$Var1 != m$Var2, ]
colnames(m) <- c("to", "from", "weight")