I'm starting to use Knime to train a Neural Network with PNN Learner and I saw that it only uses one core of CPU while running the learning. The problem is that I have a huge ammount of data (somethig like 500000 rows of input) and two parameters to predict (with 5 working as input of the Learner).
Is it possible this node run with more than one core? Or is there another way to run predictions using Parallel Chunk, for example?
Sorry if this questions are trivial, but I'm starting now.
I would say that to properly train your model you need to use the whole training set into your learner node. This means that if the node is not designed to parallelize on your different CPU cores, you cannot do it by yourself.
Conversely, once that you have your learning model trained, you can use Parallel Chunk Start and Parallel Chunk End nodes to parallelize the prediction of new items. You can do this by simply enclose your PNN predictor node by these 2 nodes.
Hope this helps,
the Parallel Chunk really works fine with the PNN Predictor, but my problem is the time spent running the PNN Learner. The last data I runned stood four days executing the Learner and didn't converge, having 50 epochs. At this level I don't know if the problem is in the data that won't converge anyway or if it is just too big and will consume so many time. I'll try to work better the input data and see what happens.
Thank you very much for your attention. Best regards,