Boris Steipe
2017-Dec-08 04:39 UTC
[R] Curiously short cycles in iterated permutations with the same seed
I have noticed that when I iterate permutations of short vectors with the same seed, the cycle lengths are much shorter than I would expect by chance. For example: X <- 1:10 Xorig <- X start <- 112358 N <- 10 for (i in 1:N) { seed <- start + i for (j in 1:1000) { # Maximum cycle length to consider set.seed(seed) # Re-seed RNG to same initial state X <- sample(X) # Permute X and iterate if (all(X == Xorig)) { cat(sprintf("Seed:\t%d\tCycle: %d\n", seed, j)) break() } } } Seed: 112359 Cycle: 14 Seed: 112360 Cycle: 14 Seed: 112361 Cycle: 8 Seed: 112362 Cycle: 14 Seed: 112363 Cycle: 8 Seed: 112364 Cycle: 10 Seed: 112365 Cycle: 10 Seed: 112366 Cycle: 10 Seed: 112367 Cycle: 9 Seed: 112368 Cycle: 12 I understand that I am performing the same permutation operation over and over again - but I don't see why that would lead to such a short cycle (in fact the cycle for the first 100,000 seeds is never longer than 30). Does this have a straightforward explanation? Thanks! Boris
Eric Berger
2017-Dec-08 05:34 UTC
[R] Curiously short cycles in iterated permutations with the same seed
Hi Boris, Do a search on "the order of elements of the symmetric group". (This search will also turn up homework questions and solutions.) You will understand why you are seeing this once you understand how a permutation is decomposed into cycles and how the order relates to a partition of n (n=10 in your case). Enjoy! Eric On Fri, Dec 8, 2017 at 6:39 AM, Boris Steipe <boris.steipe at utoronto.ca> wrote:> I have noticed that when I iterate permutations of short vectors with the > same seed, the cycle lengths are much shorter than I would expect by > chance. For example: > > X <- 1:10 > Xorig <- X > start <- 112358 > N <- 10 > > for (i in 1:N) { > seed <- start + i > for (j in 1:1000) { # Maximum cycle length to consider > set.seed(seed) # Re-seed RNG to same initial state > X <- sample(X) # Permute X and iterate > if (all(X == Xorig)) { > cat(sprintf("Seed:\t%d\tCycle: %d\n", seed, j)) > break() > } > } > } > > Seed: 112359 Cycle: 14 > Seed: 112360 Cycle: 14 > Seed: 112361 Cycle: 8 > Seed: 112362 Cycle: 14 > Seed: 112363 Cycle: 8 > Seed: 112364 Cycle: 10 > Seed: 112365 Cycle: 10 > Seed: 112366 Cycle: 10 > Seed: 112367 Cycle: 9 > Seed: 112368 Cycle: 12 > > I understand that I am performing the same permutation operation over and > over again - but I don't see why that would lead to such a short cycle (in > fact the cycle for the first 100,000 seeds is never longer than 30). Does > this have a straightforward explanation? > > > Thanks! > Boris > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/ > posting-guide.html > and provide commented, minimal, self-contained, reproducible code. >[[alternative HTML version deleted]]
William Dunlap
2017-Dec-08 16:48 UTC
[R] Curiously short cycles in iterated permutations with the same seed
Landau's function gives the maximum cycle length of a permutation. See.,e.g., http://mathworld.wolfram.com/LandausFunction.html. Bill Dunlap TIBCO Software wdunlap tibco.com On Thu, Dec 7, 2017 at 9:34 PM, Eric Berger <ericjberger at gmail.com> wrote:> Hi Boris, > Do a search on "the order of elements of the symmetric group". (This search > will also turn up homework questions and solutions.) You will understand > why you are seeing this once you understand how a permutation is decomposed > into cycles and how the order relates to a partition of n (n=10 in your > case). > > Enjoy! > Eric > > > On Fri, Dec 8, 2017 at 6:39 AM, Boris Steipe <boris.steipe at utoronto.ca> > wrote: > > > I have noticed that when I iterate permutations of short vectors with the > > same seed, the cycle lengths are much shorter than I would expect by > > chance. For example: > > > > X <- 1:10 > > Xorig <- X > > start <- 112358 > > N <- 10 > > > > for (i in 1:N) { > > seed <- start + i > > for (j in 1:1000) { # Maximum cycle length to consider > > set.seed(seed) # Re-seed RNG to same initial state > > X <- sample(X) # Permute X and iterate > > if (all(X == Xorig)) { > > cat(sprintf("Seed:\t%d\tCycle: %d\n", seed, j)) > > break() > > } > > } > > } > > > > Seed: 112359 Cycle: 14 > > Seed: 112360 Cycle: 14 > > Seed: 112361 Cycle: 8 > > Seed: 112362 Cycle: 14 > > Seed: 112363 Cycle: 8 > > Seed: 112364 Cycle: 10 > > Seed: 112365 Cycle: 10 > > Seed: 112366 Cycle: 10 > > Seed: 112367 Cycle: 9 > > Seed: 112368 Cycle: 12 > > > > I understand that I am performing the same permutation operation over and > > over again - but I don't see why that would lead to such a short cycle > (in > > fact the cycle for the first 100,000 seeds is never longer than 30). Does > > this have a straightforward explanation? > > > > > > Thanks! > > Boris > > > > ______________________________________________ > > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > > https://stat.ethz.ch/mailman/listinfo/r-help > > PLEASE do read the posting guide http://www.R-project.org/ > > posting-guide.html > > and provide commented, minimal, self-contained, reproducible code. > > > > [[alternative HTML version deleted]] > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/ > posting-guide.html > and provide commented, minimal, self-contained, reproducible code. >[[alternative HTML version deleted]]
Possibly Parallel Threads
- Curiously short cycles in iterated permutations with the same seed
- problema con los cambios de marcas temporales en el eje X
- problema con los cambios de marcas temporales en el eje X
- cdplot() with 'POSIXct' x
- [LLVMdev] Debugging Dependency Cycles between .a files for find-cycles.pl