Are there any example work flows available that use the new Similarity Search node to do something useful?
I'm trying to use it with fingerprints and it seems to work, in that it finds neighbors, but the output table doesn't seem to contain any information allowing me to join back to the table I searched from (the table on port 1, the columns from the table on port 0 are present).
Am I missing something?
In the dialog you can choose which column of the second input you want to preserve in the output. This column is then called "nearest neighbor - <RowID>". You can use it to join the output table with the second input table. It would be an inner join on this column and the rowID from the 2nd input.
We are going to polish one of the test cases and put it onto the public server.
Thanks for the answer and the sample workflow.
I have confirmed that using the Row ID as the "representative column" works, but most other columns I try produce an error like this:
WARN Similarity Search 0:10 failed to apply settings: No such column "STANDARD_VALUE" in second input.
It seems like things work ok if there's a column with the same name in the *first* table, but I don't think that's the intent of the node. :-)
Yes, not quite the intended behavior. I'll open a bug report and fix it.
.. this will be fixed in KNIME 2.7.4