Dose anyone knows why the tuple ‘cache’ be defined differently between function ‘Linear-Activation Backward’ and function ‘linear_backward’ ? The argument said that we could use ‘linear_backward’ to compute ‘Linear-Activation Backward’ but two ‘cache’ unpack the variables how to finish it?

Cache is itself a tuple, so you have to use cache[0] if you want linear_cache, or cache[1] for activation_cache.

Aha, I got it. Thanks a lot!

1 Like