Hints on compute_pca could be improved

Just finished the Week3 assignment. I had a look on hints after completing the assignment, and found out that Hints could be improved along with some descriptions in the course.

  1. The last 4 points of hints regarding transpose are redundant, it is basically the same without transpose.
    Just using matrix multiplication.
  2. It could be much clear to state out that each word is an observation with n features, which corresponding to a (nxp) data matrix, therefore, each row is a word vector.

In Deep learning specialization, Andrew prefers to use (p x n) to represent data, but here is (n x p). I think making it clear upfront is very important as this is the final data structure before applying models.

Hi Eureka,

Thanks for your comments that can be of use to other learners!

1 Like