Time Comparisons

All samplers run for 1024 iterations.

C++

X <- bench::mark(
    "Metropolis-Hastings" = {samplr::sampler_mh(1, "norm", c(0,1), sigma_prop=1)},
    "MC3" = {samplr::sampler_mc3(1, "norm", c(0,1), sigma_prop=1)},
    "Hamiltonian Monte-Carlo" = {samplr::sampler_hmc(1, "norm", c(0,1))},
    "REC" = {samplr::sampler_rec(1, "norm", c(0,1))},
    "MCHMC" = {samplr::sampler_mchmc(1, "norm", c(0,1), )},
    "MCREC" = {samplr::sampler_mcrec(1, "norm", c(0,1))},
    check = FALSE, iterations = 50
)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
knitr::kable(as.data.frame(X[,c("expression", "min", "median")]))
expression min median
Metropolis-Hastings 842.4µs 894.25µs
MC3 8.97ms 9.69ms
Hamiltonian Monte-Carlo 7.13ms 8.26ms
REC 7.14ms 8.38ms
MCHMC 52.92ms 62.77ms
MCREC 53.5ms 59.32ms

MATLAB

tests timeit
Metropolis-Hastings 6.22ms
MC3 55.13ms