during the spring we published several new R packages on CRAN. KernelICA gives an implementation of kernel ICA, and ssaBSS has several methods for (non-)stationary source separation. shapeNA allows for robust shape estimation when there are missing values in the data.
Meanwhile I am involved in 22 packages on CRAN. Below a figure about their monthly download stats as provided from the package dlstats.
So, what do you think ?