change interface ebpm_exponential_mixture to match ebnm_normal_scale_mixture

[stephens999]

comments on
ebpm_exponential_mixture <- function(x,s,m = 2, grid = NULL, seed = 123)

we would like this to match:
ebnm_normal_scale_mixture <- function(x,
s = 1,
mode = 0,
scale = "estimate",
g_init = NULL,
fix_g = FALSE,
output = output_default(),
control = NULL)

x matches so that is ok.
s =1 just needs the default
mode - probably you don't need this for now as your code assumes mode=0
change "grid=NULL" to scale = "estimate".
seed , you don't need I think so can remove (the algorithm is deterministic)
control is to be passed to mixsqp - should be easy to implement

for g_init it seems you should introduce an exponentialmix class,
similar to this code for the laplacemix class from ebnm:

#' Constructor for laplacemix class
#'
#' Creates a finite mixture of Laplace distributions.
#'
#' @param pi A vector of mixture proportions.
#' @param mean A vector of means.
#' @param scale A vector of scale parameters.
#'
#' @export
#'
laplacemix <- function(pi, mean, scale) {
structure(data.frame(pi, mean, scale), class="laplacemix")
}

Probably you shoudl use "mode" instead of mean and set it to 0 since you only
use the 0 case for now?

[zihao12]

It is updated now. I followed your advice except for the g_init. I used the class gammamix (pi, a, b) because all my code is written for gamma case as first, and I think it would be easier to extend to mixture of gamma later