"How to see the variable importance of a neural net similar to deep learning?"

annata
annata Altair Community Member
edited April 29 in Community Q&A

Hi everyone

Pls help me. I am working on rapidminer. I am using Deep learning and neutral net to build regression model. In deep learning operator, the output has weight - this port deliver the weight of the attribute with respect to the label attribute, but in neutral net we dont have it. And i want to know how to show the weight of the attribute in neutral net.

image.png

tks

An

Answers

  • RolandJones
    RolandJones
    Altair Employee

    Hi @annata,

    Unfortunately this operator doesn't support the generation of weights. I would recommend downloading the Interpretation extension, and use Generate Interpretation to generate these weights.

    Thanks,

    Roland

  • annata
    annata Altair Community Member

    Hi @RolandJones

    Thank you for replying me

    I have one more question about neutral net. So , as you know, neutral net use sigmoid for activation function and i wonder that Does rapidminer use this formula for sigmoid: f(A)=1/(1+e^-x)?

    hope you will reply

    Best

    AN

  • RolandJones
    RolandJones
    Altair Employee

    Hi AN,

    Yes, I have checked the source code and the function is listed as you describe.

    Best,

    Roland

  • annata
    annata Altair Community Member
    edited May 1

    Hi @RolandJones

    Thank you for replying me

    Can you please check the source code about normalize in neutral net, i want to know how neutral net normalize the input and the way neutral net denormalization in the ouput.

    image.png

    best

    An