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
The attribute interaction between LogPosterior and BaseCost is quite complex and unclear in places. The issue is to refactor the structure of LogPosterior so that the attributes currently copied from the likelihood class can be updated as properties instead. One way to achieve this would be to create a new parent class for meta-cost-like classes that operates on individual cost classes. One such instance would be LogPosterior, but WeightedCost could also be included.
A few goals to achieve:
Reduce the complexity of the self.parameters attribute interaction between GaussianLogLikelihood, LogPosterior and BaseCost.
LogPosterior's attributes should point to the likelihood classes, rather than copying them as needed.
LogPosterior should not inherit from BaseCost if possible, or if it does, the attributes constructed in the BaseCost parent should be deleted when the likelihood attributes are used instead (the former is really the preference here).
Feature description
The attribute interaction between LogPosterior and BaseCost is quite complex and unclear in places. The issue is to refactor the structure of LogPosterior so that the attributes currently copied from the likelihood class can be updated as properties instead. One way to achieve this would be to create a new parent class for meta-cost-like classes that operates on individual cost classes. One such instance would be LogPosterior, but WeightedCost could also be included.
A few goals to achieve:
self.parameters
attribute interaction between GaussianLogLikelihood, LogPosterior and BaseCost.Motivation
See discussion on #505.
Possible implementation
No response
Additional context
No response
The text was updated successfully, but these errors were encountered: