I think I got it all right up to getting the subset of the eigen vectors.
I get results, but they are not the expected ones.
the shapes seem to naturally produce the same shapes in " Stuck with compute_pca function"
I am taking the dot of de-meaned and the sorted eigen vectors transposed; selecting the top n components and Transposing that.
I am pretty sure that this is the line that messes up getting the right results. Can I get a point at the right direction?
Thank you
Hi eranpick,
The provided pseudocode reads as follows:
" # transform the data by multiplying the transpose of the eigenvectors with the transpose of the de-meaned data
# Then take the transpose of that product."
The sequence of steps matters here. Also, in the step preceding this, the pseudocode suggests selecting the first n eigenvectors before you perform the dot of the eigenvectors transposed and the transpose of the demeaned input (and then the whole thing transposed). To avoid bugs, it helps following this pseudocode.
Hope this helps.