NEWS.md

In ZVCV: Zero-Variance Control Variates

ZVCV 0.1.0

• Initial release
• Includes 1st - 4th order zero variance control variates
• Regularisation is based on lasso or more generally elastic net where lambda is chosen using cross-validation and the elastic net parameter is specified

ZVCV 0.1.1

• Allows for any order polynomial (with fast implementations available for polynomial orders Q = 1-4 or dimension d = 1)
• A cross validation method which chooses the polynomial order starting at 1 and going to infinity is now implemented (this new method could allow for super-root-N convergence)

ZVCV 0.1.2

• Speeding up the higher order polynomial matrix getter using C++

ZVCV 0.1.3

• Adding k-fold cross validation to select the polynomial order and doing some documentation improvements

ZVCV 1.1.0

• Added control functionals, semi-exact control functionals and approximate semi-exact control functionals

ZVCV 2.1.0

• Making Linux friendly
• Adding more checks of input arguments
• Removing duplicates in kernel methods
• Returning the estimated coefficients in zvcv
• Changing some input arguments for zvcv:
• log_weight --> log_weights
• folds_choose --> folds
• obs_estim --> est_inds and is used to specify the estimation/fitting only samples (with the remainder being used for evaluation of the integrand
• REMOVED obs_estim_choose, the option to specify the samples for each cross-validation fold. I think this level of flexibility would rarely be required and it cause confusion when est_inds is specified.

ZVCV 2.1.1

• Fixing a bug in the sample pre-processing for the special case of no duplicates + split estimation
• Other small bug fixes

ZVCV documentation built on Nov. 2, 2022, 5:17 p.m.