Why do we transpose matrices?

Hello,

In the Vectorization module, Andrew introduces matrix multiplication and mentions that we have to transpose one of the matrices to multiply the matrices. Why do we transpose matrices? Can’t we multiply them as they are?

I found an answer online that it is more efficient/easier for machines to multiply matrices by transposing, but why is it easier?

Thank you!

Sometimes you have to transpose a matrix, in order to get the rows and columns aligned with the assumptions that were made about the orientation of the weights.

There’s no universal standard for how the training data matrix is organized.

3 Likes