Create simple unit test for all methods by fitting linear regression and logistic regression models to some synthetic datasets (eg from LRVGA paper), and comparing to FCEKF and offline SGD as “gold standard”. All estimators should be able to get within epsilon predictive error given enough data.
Create simple unit test for all methods by fitting linear regression and logistic regression models to some synthetic datasets (eg from LRVGA paper), and comparing to FCEKF and offline SGD as “gold standard”. All estimators should be able to get within epsilon predictive error given enough data.