Hi Gregor,
I'm afraid in the first exercise notebook there is a slight mix-up with the dimensions of the matrices and vectors involved. In the introductory text of 1.2.2 it says and . However, if the input vector is and the output vector is , as specified in the code blocks, and following , then the weight matrix has the dimension and the bias vector is . The same applies to the text in 1.2.4, where and have been swapped.
Please let me know if I am wrong.
Best regards,
Lasse