Open kir0ul opened 1 year ago
activity[j]
and delta[j + 1]
could be converted to a matrix to remove the need to check for the dimension of the array.The code could be abstracted a bit more to higher level concepts, e.g., it could make sense to call higher level functions like a forward_pass()
, backward_pass()
, activation()
(instead of copy/pasting several times the full length sigmoid function).
Could borrow some ideas from this reference code: https://github.com/mnielsen/neural-networks-and-deep-learning/blob/master/src/network.py, taken from Neural Networks and Deep Learning by Michael Nielsen
At the moment, to convert the
Cell
data structure from Matlab, standard Python lists are used. Could be worth usingAwkward Array
to make computations faster and allow some broadcasting, see: