Different neural networks in a loop

Hi Guys,

i’am new to knime and try to use different settings for my neural networks (max nr of iterations, nr of hidden layers, nr of hidden neurons) like the guy in https://www.knime.com/forum/knime-general/scoring-different-neural-networks-in-a-loop
I’am not sure how to use the “table row to variable loop start” and how to pass the variables to my Rprop MLP Learner.

Do i have to change the column name to maybe max nr of iterations for example?
Is that loop correct?
Where can i find more Information about that type of node?

Any help would be appreciated :slight_smile:

Best regards,
Johannes

Hi @schildjo -

One strategy you could try is the use of the Parameter Optimization Loop nodes. There is a nice workflow demonstrating their use on the example server:

https://www.knime.com/nodeguide/analytics/optimization/parameter-optimization-two-examples

Hi Scott,

thanks for your suggestion.
I try to use your idea with the Parameter Optimization Loop but i’ am not sure, why i got the following error. The console says “ERROR Parameter Optimization Loop End 3:52 Execute failed: Flowvariable ‘null’ not found!” but i don’t know why. My workflow looks like this:


why the endpoint says “Flowvariable null not found”?
I compared it with the nice example from you https://www.knime.com/nodeguide/analytics/optimization/parameter-optimization-two-examples
and in the first example there is also an Scorere with the same connection and without a definition of a flow variable, so i don’t get it, what i’am doing wrong :slight_smile:
Do you have any idea?

Thanks in adanvce
Johannes

Are you able to post your workflow? If we look closely at the flow variables passed at each point in the loop, I bet we can find the answer.

Hi Scott,

thanks for your help and sorry for my late reply.
I fixed the problem by myself. Changing the Flow variable with objective function value from the Parameter Optimization Loop End did the trick :slight_smile:

Thanks again.

Best regards
Johannes