Deeplearning4J crashes KNIME 4.3

[nojrebrab]

Hi
I’ve been running a machine learning model using the MLPLearner & Predictor nodes and thought I’d try using DL4J. I created a simple model but when I tried to execute the DL4J Feedforward Learner (Regression) it processed briefly then Knime suddenly disappears from my screen. The model was similar to Deeplearning4J | KNIME. This link says there was a problem but that it was fixed in 3.3.

I’m running Windows 10 and have Anaconda Python distribution installed.

How do I get DL4J to run?

Thanks

[ScottF]

Hi @nojrebrab -

Sorry for the trouble here. I will say that we haven’t been doing much active development work on the DL4J nodes for a while now, and we recommend using the Keras nodes instead.

Here’s the one-time setup for deep learning in our documentation: KNIME Deep Learning Integration Installation Guide

And here’s an example workflow on the Hub that shows a classification model on the MNIST dataset. Look inside the “Keras KNIME Nodes” metanode to see how the individual layers are configured:

KNIME Deep Learning - Train MNIST classifier with Keras nodes – scottf

I hope that helps, but please post here if you have additional questions.

[nojrebrab]

Thanks Scott, I’ll try Keras - I would have chosen it instead of DL4J but didn’t realise it was available
John