Skip to contents

EM sweep

Source: R/fit.R
em_step.Rd

Each sweep updates every weight, then moves every atom within its own part. The support never changes size; only fw_step() or lb_step() add atoms.

Usage

em_step(state, times = 1L, record_gap = FALSE, until = NULL)

Arguments

state

A ripr_state.

times

Steps to take.

record_gap

Sweep the Frank–Wolfe oracle over the produced mixture, filling gap_after and caching it for gap_below() and fw_step(). Off by default: a full oracle sweep per row.

until

Optional predicate function(state, candidate) tested before each step; see predicates.

Value

The updated ripr_state.

See also

oracles, predicates

Examples

set.seed(1)
fam <- multinomial_family(n_trials = 4L, k = 3L)
plurality <- null_model(
  fam,
  list(
    simplex_region(vertices = rbind(c(0.5, 0.5, 0), c(0, 1, 0), c(0, 0, 1))),
    simplex_region(vertices = rbind(c(0.5, 0, 0.5), c(0, 1, 0), c(0, 0, 1)))
  )
)
Q <- fam(c(0.4, 0.35, 0.25))
state <- ripr_init(Q, plurality) |> fw_step(times = 3L) |> em_step(times = 10L)
state@trace$kl
#>  [1] 0.024511510 0.016814201 0.016233235 0.014378214 0.011977095 0.010831922
#>  [7] 0.010065905 0.009486985 0.009018320 0.008622740 0.008280092 0.007978194
#> [13] 0.007708945 0.007466560