Marc W Howard on Nostr: npub1qtkdu…txqrw So whither rotational dynamics? Briefly unless you're working on ...
npub1qtkdu9s8mnygvkcvycw0wtmp85t95v25v8eyavtkx2q0nhnuxmnq9txqrw (npub1qtk…xqrw)
So whither rotational dynamics? Briefly unless you're working on like spinal cord, you can probably ignore rotational dynamics. It's an artifact of doing linear dim reduction on diagional-ish covariance matrices like the hypothetical place field experiment above.
Faced with a diagonal covariance matrix, you can choose any basis functions you want. Most PCA algorithms end up with Fourier basis functions so the first two are like sin and cos. So you see ``rotations''.
So whither rotational dynamics? Briefly unless you're working on like spinal cord, you can probably ignore rotational dynamics. It's an artifact of doing linear dim reduction on diagional-ish covariance matrices like the hypothetical place field experiment above.
Faced with a diagonal covariance matrix, you can choose any basis functions you want. Most PCA algorithms end up with Fourier basis functions so the first two are like sin and cos. So you see ``rotations''.