Create update function for smc_mallows_new_users()

[osorensen]

I'm thinking of something like this.

First, the user runs smc_mallows_new_users() on the data they have on of these:

# one of these:
smc_init <- smc_mallows_new_users(...)
smc_init <- smc_mallows_new_item_rank(...)

If the S3 class of smc_init tells whether they are from the "new users" or "new item rank" case, then we can have a common update function smc_mallows_update. When they get new data, they do this:

smc_updated <- smc_mallows_update(smc_init, ...)

[osorensen]

For smc_mallows_new_users(), I think the initialize phase can be easily done in R, and then we're left with a loop over new users in the C++ code:

BayesMallows/src/smc_mallows_new_users.cpp

Lines 81 to 112 in 57bb333

	/* ====================================================== */
	/* Initialise Phase */
	/* ====================================================== */
	int n_users = R_obs.n_rows; // total number of users
	if(type == "complete"){
	if (Time > n_users / num_new_obs) {
	Rcpp::warning( \
	"Time should not exceed n_users / num_new_obs. Recalculating."\
	);
	Time = n_users / num_new_obs;
	}
	}
	/* generate rho samples using uniform prior ------------- */
	cube rho_samples(N, n_items, Time + 1);
	rho_samples.slice(0) = initialize_rho(n_items, N).t();
	/* generate alpha samples using exponential prior ------- */
	mat alpha_samples;
	if(type != "partial_alpha_fixed") {
	alpha_samples = zeros(N, Time + 1);
	alpha_samples.col(0) = initialize_alpha(N);
	}
	/* generate vector to store ESS */
	rowvec ESS_vec(Time);
	// this is to store the augmentations of the observed rankings for each particle
	cube aug_rankings; // no. users by items by particles
	if(type == "partial" || type == "partial_alpha_fixed"){
	aug_rankings = zeros(n_users, n_items, N);
	}