Clears the cached numerical derivatives (score and Fisher) from an mle_problem. This is useful when you want to force recomputation, for example after modifying data that the log-likelihood depends on.
Examples
if (FALSE) { # \dontrun{
problem <- mle_problem(loglike, cache_derivatives = TRUE)
# ... run some optimization ...
clear_cache(problem) # Force fresh derivative computation
} # }