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
Me and @rubiop seem to be tackling problems where we independently need functionality for two kinds of maps
Fixing a slice of the input variables and pre-evaluating the corresponding functions, at least the beginning of the map (e.g. we have T(x,y) and we want the map T_y(x*,•), which is square)
Taking the inverse of a map and using it as a map in and of itself (e.g. we have a map x = T(y) and we want the map y=T^{-1}(x) ). A lot of functions like Evaluate and Inverse are straight forward, Some functions like CoeffGrad are tricky, and some like SetCoeffs are extremely unclear.
The text was updated successfully, but these errors were encountered:
InverseMap is really hard-- the gradients become ridiculous (you end up having to do three-dimensional tensor contractions of derivatives with order >1)-- so we're definitely not adding that any time soon.
Me and @rubiop seem to be tackling problems where we independently need functionality for two kinds of maps
Evaluate
andInverse
are straight forward, Some functions likeCoeffGrad
are tricky, and some likeSetCoeffs
are extremely unclear.The text was updated successfully, but these errors were encountered: