How does increasing the neural networks layers (making it bigger) help reduce the bias?

I can understand that reducing regularization parameter, adding more features can help reduce the bias and how they do so. But i can’t grasp how a bigger neural network deals with bias

Bigger NN’s can learn more complex models. So that reduces bias (i.e. they’re more likely to overfit rather than underfit).


Hello @Jay_01,

As though reducing regularization parameter and adding more features can help with reducing the bias, but in the other hand, it could increase the chance of overfitting.
So as @TMosh mentioned, bigger neural network could lead to smaller bias but the efficiency of the neural net and knowing the problem is more important than just adding layers.


1 Like

That was silly of me, thank you for your explanations.