Here’s a graph in which nodes (and edges) represent currencies (and exchange rates):

library(igraph)
currencies <- factor(c("EUR", "USD", "JPY", "GBP"))
df <- subset(expand.grid(from=currencies, to=currencies),
             from != to)
GetExchangeRates <- function(from, to) {
  urls <- sprintf("%s/d/quotes.csv?s=%s%s=X&f=b",
                  "http://download.finance.yahoo.com", 
                  from, to)
  GetRateFromUrl <- function(str) {
    message("Reading from ", str)
    tryCatch(read.csv(url(str), header=FALSE)[1, 1],
             error = function(e) NA)
  }
  sapply(urls, GetRateFromUrl)
}
# If a url connection fails, the corresponding rate will be NA
df$rate <- GetExchangeRates(df$from, df$to)
g <- graph.data.frame(df, directed=TRUE)
g$layout <- layout.fruchterman.reingold(g)
E(g)$label <- E(g)$rate
V(g)$label <- V(g)$name
dev.new(width=10, height=10)
plot(g, main=sprintf("Exchange Rates on %s", Sys.Date()))
savePlot("exchange_rate_graph.png")

I'd like to emulate this post and look for profitable cycles using R. Here's a first attempt:

# Look for negative-cost cycles
E(g)$weight <- -log(E(g)$rate)
shortest.paths(g)

In this case, the shortest.paths function complains that it "cannot run the Bellman-Ford algorithm" because a "negative loop [was] detected while calculating shortest paths" -- great! There's a negative-cost cycle in there somewhere. But what's the easiest way to actually find that cycle using R -- does anyone have any tips?