(Choose 1 answer)
After training a neural network with Batch Norm, at test time, to evaluate the neural network on a new example you should:
A. To ensure consistency in evaluating a test example after applying Batch Normalization on mini-batches of 256 examples during training, you would need to replicate the test example 256 times. This way, you can work with a mini-batch that matches the training conditions.
B. Use the most recent mini-batch's value of µ and σ^2 to perform the needed normalizations.
C. Perform the needed normalizations, use u and o^2 estimated using an exponentially weighted average across mini-batches seen during training.
D. Skip the step where you normalize using u and o^2 since a single test example cannot be normalized.
Exit 13
