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
just an idea for possible performance improvements while staring at some linear algebra
_smw_solve is called very often to invert the residual covariance matrix
The calculations depend on exog_re
In some special cases like random constant the exog_re are the same across all units or only depend on the length for the individual exog_re.
_smw could be converted into a class that has a multiply/whiten method and keep the repeated calculations around. (similar to chol_factor, chol_solve or others)
just an idea for possible performance improvements while staring at some linear algebra
_smw_solve
is called very often to invert the residual covariance matrixThe calculations depend on exog_re
In some special cases like random constant the exog_re are the same across all units or only depend on the length for the individual exog_re.
_smw
could be converted into a class that has a multiply/whiten method and keep the repeated calculations around. (similar to chol_factor, chol_solve or others)related to #1948
The text was updated successfully, but these errors were encountered: