Remove Bias from neural net

19 vues (au cours des 30 derniers jours)
Andriy Artemyev
Andriy Artemyev le 20 Mai 2021
Commenté : Andriy Artemyev le 24 Mai 2021
Greetings, I have a neural network that is forecasting financial data. In some cases, my 3 input features are going to be full of zeros. In those cases, my network will give me a nonzero output, even if all the inputs are zero. This is very problematic for my particular problem. I think this is due to the biases in the network. Is there a way to train a networks weights, but force the biases to zero? I also fitted a linear regression model, where I could simply remove the intercept to solve this problem.
I unfortunately could not find information on this topic and hope that someone can help me out!
Thanks in advance,
Andy

Réponse acceptée

Srivardhan Gadila
Srivardhan Gadila le 22 Mai 2021
You can remove the bias by setting the value of net.biasConnect of all the layers of the neural network to zero. Refer to the documentation of net.biasConnect for more information. If you want to access the bias value of the neural network refer to the documentation of net.b.
  1 commentaire
Andriy Artemyev
Andriy Artemyev le 24 Mai 2021
Thank you very much I will try this out!

Connectez-vous pour commenter.

Plus de réponses (0)

Catégories

En savoir plus sur Sequence and Numeric Feature Data Workflows dans Help Center et File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by