I’m doing the second programming assignment: “regularization” I’m confused as to why A1 and A2 are multi dimensional matrices. I thought the the activations were meant to be of shape (X, 1), representing one layers of activation in the neural network.

The second (column) dimension of the activation matrices (A1, A2, …) is the number of samples, right? The point is that we are vectorizing forward propagation across the samples (in the batch or mini-batch as appropriate).