Skip to content
Interview prep / ML/Data

ML/Data interview questions: page 3 of 3

Questions 101–105

  1. Why nonlinear activations matterEasy
  2. All-zero hidden-layer initialisationMedium
  3. Dropout when making predictionsEasy
  4. Unstable loss after a learning-rate increaseMedium
  5. Training and evaluation modes for normalisationMedium