TODO.md
In nhejazi/haldensify: Highly Adaptive Lasso Conditional Density Estimation

`haldensify` development notes

[x] Add ... args to haldensify() to allow arbitrary arguments to be passed directly to fit_hal().
[x] Remove use_future argument to haldensify(), instead reducing to calling future_mapply(), with sequential evaluation via plan(sequential).
[x] Add example of working with bootstrap samples, e.g., using rsample.
[x] Add a plot() method to more easily visualize how empirical risk changes across the sequence of explored regularization parameter values.
[x] Add an ipw_shift() function for constructing IPW estimators of the mean counterfactual outcome of a stochastic shift intervention via haldensify.
[ ] Add an argument to haldensify() to allow normalization of the density estimates to improve estimation stability. Note that this normalized density is actually g(A|W)/g(A), instead of the currently estimated g(A|W).