Changes in version 1.4.1 (2025-05-19) BUG FIXES o bugfix to avoid valgrind error (Koen Derks) o small bug fix by Max Kuhn Changes in version 1.4.0 (2025-04-16) OTHER CHANGES o removed deprecated function simulation o man pages are created with roxygen2 now o some bug fixes Changes in version 1.3.2 OTHER CHANGES o added native routine registration o added some unit tests BUG FIXES o checks that k is smaller than the number of observations in the learning set o if k was equal or larger the number of observation in the learning set result were spurious Changes in version 1.3.1 (2016-03-26) OTHER CHANGES o predict got an additional argument "type" o changed igraph code to version >=1.0 o train.kknn has a new argument ks, a vector specifying values of k Changes in version 1.3.0 (2015-07-09) OTHER CHANGES o kknn is now hosted on github o added travis-ci support Changes in version 1.2-5 (2014-02-11) BUG FIXES o a browser() sneaked into kknn Changes in version 1.2-4 (2014-02-11) NEW FEATURES o new function kknn.dist OTHER CHANGES o kknn returns now an index matrix of nearest neighbors Changes in version 1.2-3 (2013-12-16) BUG FIXES o Namespace changed Changes in version 1.2-2 (2013-03-13) NEW FEATURES o kknn has new parameter scale to switch off scaling o new function cv.kknn for k-fold cross validation, replacing the function simulation in future o default kernel changed to optimal BUG FIXES o interaction terms in formulas should be handled correctly OTHER CHANGES o speeded up kknn code a bit and reduced some memory allocations Changes in version 1.2-1 (2012-10-30) BUG FIXES o solved some namespace problems Changes in version 1.2-0 (2012-10-05) NEW FEATURES o added new kernel "optimal" to kknn Samworth (2012). Changes in version 1.1-2 (2012-06-22) BUG FIXES o changed package to work with igraph >= 0.6 Changes in version 1.1-1 (2011-10-12) OTHER CHANGES o computation of Eucledian distances is 3-5 times faster. Changes in version 1.1-0 NEW FEATURES o added spectral clustering function specClust. OTHER CHANGES o computation of knn distances is much faster. Changes in version 1.0-8 (2010-08-05) BUG FIXES o the data.frame for the training data set does not have to contain the response variable (suggested by Carlos J. Gil Bellosta). Changes in version 1.0-5 (2007-10-11) BUG FIXES o A bug was fixed, leave-one-out crossvalidation now works fine. Changes in version 1.0-3 NEW FEATURES o new parameter ykernel: Window width of an y-kernel, especially for prediction of ordinal classes. BUG FIXES o A bug was fixed in the kernel="rank" in kknn. o kernel = "gaussian" did not work correctly: this has been fixed.