AutoML Component via Interactive Views

This application is a simple example of AutoML with KNIME Software for binary and multiclass classification. The key of this example is the AutoML Component verified and developed by the KNIME Team. This Component (in green) uses the extension KNIME Integrated Deployment in order to train several models and combine them with other nodes and output a deployment workflow. The other Components (in blue) offer interactive views to control the AutoML process from any web browser via the KNIME WebPortal on KNIME Server. The workflow also works locally on KNIME Analytics Platform. Make sure to use "Apply and Close" in bottom-right corner of each view.


This is a companion discussion topic for the original entry at https://kni.me/w/ZBbj2fObfRwrp8ht
1 Like

Hello there,
Make sure your KNIME Analytics Platform is 4.2.1 or higher when using this workflow.
Cheers
Paolo

This is fantastic. I am sure many like me will appreciate the effort that went into making this component. These kind of components are a true time saver.
Out of curiosity, does any of the models listed in this component make use of GPU? specifically CUDA cores?

1 Like

Hi @smurari,
glad you like our work. It is indeed a time saver!
In the Component there is a Keras network (deep learning) which will use the conda enviroment you can specify you KNIME settings. The default would be to use CPU for Keras as well but I believe this can be changed there. You can check the documentation below (GPU / Cuda included). Once that is configured the AutoML Component should start using your GPU as well (for Keras training only).

docs.knime.com/2018-12/deep_learning_installation_guide

have you checked out the new XAI View?

ezgif-3-43cacf89a4ba

Workflows:

  1. https://kni.me/w/5xfwkuVsF6Uz8hMC
  2. https://kni.me/w/JZbuUdhGKZBEpdoK

Thank you @paolotamag . XAI View is another awesome component for ML workflows. Thanks for pointing me towards it.

1 Like