You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Basically the reason you can't index ExplictTangentBundles with TaylorTangentIndex except at the last position is because there are multiple possible ways to compute a given deriviatve of any order -- except the highest order, then you need to do all the things so there is only one way.
In a good world they are all equal but we may not live in a good world.
If we do live in a good world though we can.
And the thing to do is when given a TaylorTangentIndex check all values that correspond to different paths to get a derivative of that order and if all equal return that value. If not all equal then error as we do not live in a good world.
The canonical index positioned to check are the ones that have count_ones(ii) equal to the taylor index (the order of the derivative).
Computing this is apparently a well known problem, see this thread on discourse.
Probably for performance reasons want to hard code the indexs to check for order 1 and 2 (maybe 3), then fall back to one of those algorithms.
The text was updated successfully, but these errors were encountered:
Basically the reason you can't index
ExplictTangentBundle
s withTaylorTangentIndex
except at the last position is because there are multiple possible ways to compute a given deriviatve of any order -- except the highest order, then you need to do all the things so there is only one way.In a good world they are all equal but we may not live in a good world.
If we do live in a good world though we can.
And the thing to do is when given a
TaylorTangentIndex
check all values that correspond to different paths to get a derivative of that order and if all equal return that value. If not all equal then error as we do not live in a good world.The canonical index positioned to check are the ones that have
count_ones(ii)
equal to the taylor index (the order of the derivative).Computing this is apparently a well known problem, see this thread on discourse.
Probably for performance reasons want to hard code the indexs to check for order 1 and 2 (maybe 3), then fall back to one of those algorithms.
The text was updated successfully, but these errors were encountered: