How would you handle the vanishing gradient problem and exploding gradient problem?

Question

Anonymous · Accepted Answer

Use proper weight initialization (like He/Xavier), normalized architectures
(BatchNorm), and stable activations (ReLU) to prevent vanishing gradients.
Control exploding gradients with gradient clipping, smaller learning rates, and
architectures like LSTM/ResNet that preserve gradient flow.

Anonymous · Answer

In this step we want to emphasize on techniques for dimensionality reduction. We
can leverage reLU function and LSTM using sinh or tanh activation functions to
overcome exploding gradients.

How would you handle the vanishing gradient problem and exploding gradient problem?

Practice this question with AI

Go Premium

Community Answers

Unlock Full Access

Unlock Full Access

Practice More Questions