Machine Learning — Medium
Key points
- Gradient clipping sets a limit on gradient values to prevent them from becoming too large.
- This helps avoid numerical instability and ensures smoother training.
- It is particularly important in RNNs where gradients can explode during backpropagation.
Ready to go further?
Related questions