Simple Example for Multiclass Classification with Keras

This workflow trains a fully connected feedforward neural network with 4-8-3 units per layers to classify iris flowers.

This is a companion discussion topic for the original entry at

Reference to “Keras Network Learner Node”, I guess the correct training settings in the first tab, the Input Data tab, should be the four numerical inputs (sepal & petal lenght/width). However, in your uploaded workflow the conversion was “From Collection of Number to One-Hot Tensor” and the input was “class_collection”!

My understanding you did not apply nominal feature encoding on the input features. Moreover the selected input column was the “class_collection” which is the target!

Thanks for your kind feedback

1 Like

Hello @YasserSoliman,

thank you for pointing this out!

You are absolutely right and I will correct the workflow on the Hub


hi, I just bought the knime book weeks ago, I encounter a problem with keras network learner, everything prior to this node was executing fine, when I execute keras network learner I got this message of (ERROR Keras Network Learner 0:74 Execute failed: (“NullPointerException”): null), I don’t know what to fix, I checked my extensions were fine, please would you help me on this issue, thanks.

1 Like

Any solution for this yet? get the same error.