Added documentation for nearest-neighbor-gradient [cont] #169
Conversation
|
"There are three main variants of mapping" should become four I think. |
| When using the `rowsFirst` parameter, the following format is required: | ||
|
|
||
| ```cpp | ||
| ( v0x_dx, v0y_dx, v1x_dx, v1y_dx, ... , vnx_dx, vny_dx, | ||
| v0x_dy, v0y_dy, v1x_dy, v1y_dy, ... , vnx_dy, vny_dy ) | ||
| ``` |
There was a problem hiding this comment.
This goes beyond the documentation part, but considering that the 'gradient matrix' is a Jacobian, I think the rowFirst argument should be a column first, as the Jacobian contains the derivative in a particular direction in a column, right?
There was a problem hiding this comment.
Jacobian contains the derivative in a particular direction in a column, right?
yes.
But here, we have a 3D tensor, right? So "rowFirst" or "columnFirst" will never tell the full story, right?
| When using the `rowsFirst` parameter, the following format is required: | ||
|
|
||
| ```cpp | ||
| ( v0x_dx, v0y_dx, v1x_dx, v1y_dx, ... , vnx_dx, vny_dx, | ||
| v0x_dy, v0y_dy, v1x_dy, v1y_dy, ... , vnx_dy, vny_dy ) | ||
| ``` |
There was a problem hiding this comment.
Jacobian contains the derivative in a particular direction in a column, right?
yes.
But here, we have a 3D tensor, right? So "rowFirst" or "columnFirst" will never tell the full story, right?
Co-authored-by: Benjamin Uekermann <benjamin.uekermann@gmail.com>
|
This is now ready to merge once #169 will be merged. |
Follow-up to #166