I have a table whose columns I would like to normalize in my KNIME workflow. I have found that for this particular table if I try to use the "Normalizer" node, it sees no columns in the table and will not run. However if I take the exact same table (which is output from the step prior) and instead send it to "Normalizer (PMML)" it sees the columns and will normalize my data without incident.
Can someone help me understand the differing behavior between these nodes?
Can you give a bit more information. What are the data types of the input columns, and which KNIME Analytics Platform version are you using?
Thank you for suggesting that I look at the input column types - it turns out that due to earlier manipulations of the table they had a type of "?". By assigning types (I opted for "Column Auto Type Cast" which correctly assigned all the columns of interest to type "D") I was able to then use the "regular" (non-PMML) normalizer.
Is this a feature on one or the other than the PMML normalizer can take a table with columns that are not explicitly defined as a numeric type while the non-PMML normalizer cannot?
This is KNIME 3.5.1 in windows server 2k12r2.
I checked this on my installation and both normalizers do not show columns that have a question mark. However, it may depend on how this unknown type came to be in the first place. Could you provide me with the workflow where this problem occurs?