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
I wonder if rather than this package defining impute and impute_missing it would be instead be worth implementing the interface provided by Impute.jl?
One possible way could be to define a LowRank <: Imputor and extend impute(A, imp::LowRank)
along the lines of
struct LowRank <:Impute.Imputor
losses::Vector{Loss}# Vector of loss functions
rx::Vector{Regularizer}# Vector of regularizers to be applied to each column of X
ry::Vector{Regularizer}# Vector of regularizers to be applied to each column of Y
k::Int# Desired rankendfunction Impute.impute!(A, imp::LowRank)
# fit glmr, to get  = X Y' ≈ A# If A[i, j] is `missing`s replace it with the value Â[i, j]endfunction Impute.impute(A, imp::LowRank)
# fit glmr, to get  = X Y' ≈ A# return Âend
This is a great idea. I wouldn't do this instead of but in addition to the methods defined internally, in a separate file. (It's better not to be too dependent on dependencies.)
I'd welcome this as a PR; it would be an easy project for someone looking to contribute to this package.
I wonder if rather than this package defining
impute
andimpute_missing
it would be instead be worth implementing the interface provided byImpute.jl
?One possible way could be to define a
LowRank <: Imputor
and extendimpute(A, imp::LowRank)
along the lines of
Here are examples of current imputors :)
The text was updated successfully, but these errors were encountered: