How to skip a loop optimization step if an error occurs?

KNIME Analytics Platform
1

Dear KNIMErs,

I'm using parameter optimization loop nodes in order to optimize the parameters of a SVM learner. One of the SVM kernel types, which is between the optimizing parameters, give me the following error:

ERROR SVM Learner Execute failed: No support vectors could be found for class Class B. Consider using a different kernel.

After that the error is triggered the optimization loop stops. Is there a way in order to skip an optimization step and follow with the other loop parameters if such an error occurs during the optimization? I'm using a brute force optimization strategy.

Gio

2

I'd use a Try-Catch node combination inside the loop.

3

Thanks for the suggestion Thor. I wasn't aware of such a nodes.

Is it possible that the Try-Catch node combination cannot be used in a Cross Validation loop? I'm having the following errors:

​ERROR     SVM Learner                        Execution failed in Try-Catch block: No support vectors could be found for class Poor absorption. Consider using a different kernel.
ERROR     X-Aggregator                       Execution failed in Try-Catch block: No matching Loop Start node!

I also uploaded a pict of my workflow. Or maybe I'm using it in a bad way?

Gio

4

I wouldn't recommend a try and catch part in and part out of a loop personally. 

Try moving the Try node Inbetween the X Partitioner node and SVM Learner node, see if that helps.

simon.

5

OK Simon,

Thank you for the suggestion. I didn't have experience with the try-catch construct. Now it works.

Cheers

Gio