EasyDeepLearn

What is 'dying ReLU' and how do you fix it?

medium

Answer

  • A ReLU neuron that always outputs zero has a zero gradient everywhere and stops learning permanently.
  • Causes: large negative bias, extreme learning rate, poor init.
  • Fixes: Leaky ReLU or ELU (nonzero slope on the negative side), He initialization, lower learning rate, better data normalization, or use GELU / SiLU which are smooth and don't fully saturate to zero.
Check yourself — multiple choice
  • Dying ReLU only affects the final layer
  • A neuron stuck at 0 with zero gradient — fix with Leaky/ELU/GELU, He init, or lower LR
  • It resolves by itself
  • Only dropout can fix it

Zero output → zero gradient → permanently dead. Leaky/ELU/GELU + good init fix it.

#activations#training

Practise Deep Learning

214 interview questions in this topic.

Related questions