Understanding Neuronal Networks

Hi all,

How can be Neuronal Networks understood? Once I create the workflow, which is the best way to display results and what kind of results might be of interest when working with NN?

Any tip is welcome!

Thanks in advance,


The neural network (the model) itself is usually hard to interpret. You usually want to use this to apply test or validation data on the model and evaluate the performance using ROC or Lift charts. Hope this helps to get started.

Thanks Gabriel, will look for that. I guess also studying the MSE after a X-validation is useful, to see how accurate is your model.


Hi everyone, 

I want to know more about the ANN implementation in Knime. I've read some of the documentation but I'm still a little bit confuse.

Which is the stop function? I mean, I think that, despite I split my database into training and testing set before the net, inside the learning ANN node itself there is another split, between training and validation, right? If we don't have this, we could arrive to an overfitting problem with the test set in the next prediction ANN node, if only the optimization of the training set error is the goal in the learning node. 

If I'm wrong I appreciate any help someone could give me :)

Have a nice day!