Predefine colors for Color Manager

I train a decision tree and find out which of my rows have been correctly classified and which wrongly. Sometime everything is “right”, sometime some rows is “right” and some are “wrong” and sometime everything is “wrong” if I train on some extremely small set (e.g. with 3 rows, just to test technical functionality of the flow). Some of the rows are also labeled “train”, as I do not classify my training data (using a Partitioning node before the training and testing). I then parse this column “correctness” to the Color Manager node.

Long story short (TL;DR): I have a String column “correctness” with varying entries for different datasets.

Every time the Color Manager gets different values for “correctness” I need to open it and press “OK”. This is fine on my personal machine, but once I upload the flow to my KNIME Server, this does not work anymore - the flow “does not fully execute”.

Question: How can predefine for the Color Manager what values it should expect and which colors the respective value/rows should have if a particular value is encountered in the “correctness” column?

Hey M42,
the Color Manager auto-guesses a column if it gets connected to another node. Then it looks at the values in the nominal column and colors them in the RGB color space. You can find the values of the nominal columns in the table spec view (2nd tab in table output).

You can change those values with the “Edit Nominal Domain” Node. So if you enter 3 values there “wrong, right, train” then the values will always be present in the domain and the Color Manager will color your values in THAT order even if values are not present in your dataset. Unfortunately, there is no real way to specify the exact color, but if you enter “wrong”, “right”, “train” in the Edit Nominal Domain your wrong will be colored red, right green and train blue.

I hope I could answer your question.