add jitter to training data

can anyone please suggest how to add 5% or 10% error to the training data and make the model again to see the sensitivity of the model to noise ?? thanks

Hi @abhisek -

As a first pass, you might try out the Stresser and/or Gaussian Distributed Assigner nodes. Here’s an example workflow from the Hub that highlights both nodes. (Also, here’s a forum thread that provides more info on how the Stresser works.)

Of course if you want very fine tuned control, then you might consider using the Math Formula node to calculate noisy data manually.

1 Like

thanks for your suggestion…:slight_smile:

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.