Which of the following techniques can help mitigate the vanishing or exploding gradients problem in deep neural networks? Select all that apply.

Mitigating vanishing or exploding gradients can be done by using a good weight initialization scheme such as Xavier or He initialization, which keeps the variance of activations and gradients stable across layers. Gradient clipping limits the maximum value of gradients during back‑propagation, preventing them from blowing up. Batch‑normalization normalizes layer inputs, reducing internal covariate shift and keeping gradients in a reasonable range. Residual connections (skip connections) allow gradients to flow directly to earlier layers, bypassing the chain of multiplications that cause vanishing gradients. Using activation functions with non‑zero gradients like ReLU or its variants also helps keep gradients from dying.