To perform linear correlation filtering one first needs to run the “Linear Correlation” node. The “model” of this node can then be used in the “Correlation Filter”.
The issue is that “Linear Correlation” node is slow and it becomes very, very slow with large number of columns. Just doing it in python is a lot faster.
It would be helpful to have similar speed on the KNIME native solution and not having to reinvent the wheel. But it’s more than 10x times slower, in my case python is 3.6 seconds while the 2 knime nodes take 133 seconds! So there is a lot of room for improvement.