I use a multi layer perceptron to generate my regression model.
I want to show how the error decreases in the interations, to show the elbow etc.
I see no possiblility to do that, not even in the weka mlp node. Also no possibility to set the paramters like traing time etc within a loop (over flow vars). That would be nice for optimization. Any ideas?
The Rprop node has a view available through the node context menu that shows the error in each iteration in a scatter chart. Hope this helps. However, this node (but not the Weka node) can also embedded into a loop context running it for different numbers of iterations.
Ah okay thats true. Thanks for that hint. Haven't thought of working with the RProp implementations.
However I think it is not possible for the weka plugins. I would preper to use these nodes. In the RProp it is not even possible to set a learning rate.