Inverted droput C2 Week 1

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
2

The common mistakes are:

  1. Using the wrong random function to compute the mask. We need numbers evenly distributed between 0 and 1, so using np.random.randn will end badly.
  2. Creating the mask as a column vector instead of a matrix the same shape as A. See this thread for more discussion of that point.
  3. Forgetting to do the “inverse” calculation after applying the mask.