What is inverted dropout?

Question: What is the idea behind inverted dropout? A) Drop all weights permanently after training B) Increase dropout probability during inference C) Normalize labels before applying dropout D) Scale activations during training so no extra scaling is needed at inference Correct: D Explanation: In inverted dropout, surviving activations are scaled during training. This keeps the expected activation magnitude consistent, allowing dropout to be disabled at inference without additional output scaling. Topic: advanced ML / deep learning / dropout