| Factorization with noisy
data |
|
|
|
|
|
|
|
| Once again: use SVD of W |
||||||
| • | Set all but the
first three singular values to 0 |
|||||
| • | Yields new matrix
W’ |
|||||
| • | W’ is
optimal rank 3 approximation of W |
|||||
|
|
|
|
|
|
|
| Approach |
||||||
| • | Estimate W’,
then use noise-free factorization of W’ as before |
|||||
| • | Result minimizes
the SSD between positions of image features |
|||||
| and projection of
the reconstruction |
||||||