bgoodri
bgoodri
(1) I imagine it will end up like the ODE solvers where we have slightly different versions with a mostly common set of arguments. But Halley is probably the one...
(2) It seems like it would be good to permit (or require) one functor that returns a tuple of: level, first derivative, second derivative.
(3) If a root needs to be found for multiple parameter values, I would say that the caller is responsible for invoking `root_finder` multiple times.
I think we should default the `digits` argument according to the recommendation in the Boost [docs](https://www.boost.org/doc/libs/1_79_0/libs/math/doc/html/math_toolkit/roots_deriv.html). > The value of digits is crucial to good performance of these functions, if...
You could also easily test `a * x^2 + b * x + c = 0` where `a`, `b`, and / or `c` are `var`, `fvar`, etc. to make sure...
From your email, `guess`, `min`, and `max` should definitely be templated but we just need to pull out the underlying numbers since we wouldn't need to differentiate with respect to...
Also, @charlesm93 might have some thoughts. As I mentioned yesterday, the motivation for having a one-dimensional specialization of `algebraic_solver` in Stan Math (we are not worried about exposing it to...
We should probably also do the beta inverse CDF as a test case. Boost has a more advanced way of obtaining an initial guess in this case, but they still...
There are analytical derivatives for the beta quantile case at https://functions.wolfram.com/GammaBetaErf/InverseBetaRegularized/20/01/
Thanks @charlesm93