I wonder, why to define initially the matrix W to have neurons as column vectors, only later to keep transposing it all the time?
And the in the later lectures, we adopt the notion that W has its neurons as row vectors, and then we do not use transpose anymore.
Just trying to understand the initial motivation of having W as matrix with column vectors.
As you can see X is a matrix of size (n_x, m), where n_x is the number of features and m is the number of examples. In order to carry out a matrix multiplication (dot product operation) when computing Z, W will have to be transposed. Here is link that explains matrix and different mathematic operations
yes, I understand that, I wonder why not to define W to be (m, n_x) in the first place?
Hi @Oleksandra_Sopova ,
(m, n_x, m) is a 3-dimensional matrix, why do you want to do that?
Depending on the problem we are trying to solve, not every dataset is nicely organised to suit our particular problem. So, knowing how to make changes is important. Later on in the course, you will see ‘m’ is organised in rows and not in columns, here m is the number of examples.