some Isomap procedural tweaks

[topepo]

• Added a verbose option and set to TRUE. This will be good once these get wrapped inside of resampling routines
• Added a cap on the number of neighbors. Otherwise, there is an error:

  Error in RANN::nn2(data = x, query = x, k = k + 1, treetype = treetype,  : 
  Cannot find more nearest neighbours than there are points

Fo the apply function, where you rbind the out of sample data with the training set, I used

lknng <- makeKNNgraph(rbind(indata, orgdata),
                      k = min(norg, pars$knn), eps = pars$eps)

but that may not be the best idea (I have no idea).

[gdkrmr]

I don't fancy adding more parameters for printing, most methods already have too many and I don't have access to the internals of the methods that I did not implemented myself. There is a generic way to suppress the output of every method by setting .mute = c("message", "output") in embed. Including "message", suppresses normal print output, "output" suppresses error messages, warnings and messages, the exact workings are not entirely clear to me, you can check out ?sink for more information. I should make this more clear in the documentation of embed.

emb <- embed(loadDataSet("Iris"), "Isomap", .mute = c("message"))

[topepo]

In the Rd files is is minus the dot:

emb2 <- embed(dat[samp], "Isomap", mute = NULL, knn = 10)

Should I delete the PR?

[gdkrmr]

keep the part that fixes the bug with the nearest neighbors and add the fix for the documentation. Thanks by the way for checking all my code!

[topepo]

No problem; thanks for the package

[gdkrmr]

i am closing it,

• The RANN package does the checking for the number of neighbors and gives a useful error message, therefore no extra code for error checking needed.
• I am a friend of failing programs and not helping the user silently: If people put more neighbors than point in their code then probably there is something wrong and the user should be made aware of it, in Isomap one should use k << n, else one could do a simple MDS or PCA anyways.