An iterative feature selection method that internally utilizes varius Machine Learning methods that have embeded feature reduction in order to shrink down the feature space into a small and yet robust set.
The following warning is thrown when the code is run in parallel.
1: In `[<-.data.frame`(`*tmp*`, , tmp.columns.to.zscore, value = list( :
provided 114 variables to replace 25 variables
The issue is that in some unlucky cases there are only one column left while we are binning the columns to send to each CPU core. In those cases, selection of columns from a matrix object does not return a table but rather a numeric vector. This means that on the line 677 of the following chuck, we are sending a vector to our func.zscore function rather than a matrix.
The x[, tmp.bins[[i]]] in line 677 of the chunk above is equivalent of:
base::getElement(object = x, name = tmp.bins[[i]])
The solution is pretty obvious, using base::subset(), although I think it will add some overhead on the runtime of the code. But because it is run only once per each SIVS run, it is pretty okay. The fix will come in PR #8.
The following warning is thrown when the code is run in parallel.
The issue is that in some unlucky cases there are only one column left while we are binning the columns to send to each CPU core. In those cases, selection of columns from a matrix object does not return a table but rather a numeric vector. This means that on the line 677 of the following chuck, we are sending a vector to our
func.zscorefunction rather than a matrix.https://github.com/mmahmoudian/sivs/blob/7e629e2212f96106b6b5fdec2eca07e255a03840/R/sivs.R#L665-L678
The
x[, tmp.bins[[i]]]in line 677 of the chunk above is equivalent of:The solution is pretty obvious, using
base::subset(), although I think it will add some overhead on the runtime of the code. But because it is run only once per each SIVS run, it is pretty okay. The fix will come in PR #8.