Last answered:

09 Apr 2020

Posted on:

08 Apr 2020

0

Strange matrix multiplication in TensorFlow2 lecture multiple inputs

video: Deep Learning with TensorFlow 2.0 > Neural networks Intro > The linear model. Multiple inputs and multiple outputs
screenshot
https://imgur.com/a/sEH7Gr0

The matrix multiplication shown around time 2:58 is not correct. It's not how matrices are multiplied. Try creating them using np.array, the output values will not be achieved.
I think the source of the confusion comes from reversing the order of the matrices. In the video, the weights matrix is on the right of the variables matrix, which is opposite what is shown the lectures notes, page 8.
Thanks

 

PS I am not referring to the machine way of adding the bias matrix, which of course is not the same as ordinary matrix addition.

1 answers ( 0 marked as helpful)
Instructor
Posted on:

09 Apr 2020

0
Hi Andrzej, Completely correct! In fact, after checking our files we found that this is a known mistake which was corrected, but we never uploaded the video. We are very grateful that you took the time to write this comment! We have now replaced the video. Best,
Iliya

Submit an answer