The {trackopt} package tracks parameter values,
gradients, and Hessians at each iteration of numerical optimizers in
R. This can be useful for analyzing optimization progress,
diagnosing issues, and studying convergence behavior.
You can install the released package version from CRAN with:
install.packages("trackopt")The following example tracks nlm while it minimizes Himmelblau’s
function:
library("trackopt")
himmelblau <- function(x) (x[1]^2 + x[2] - 11)^2 + (x[1] + x[2]^2 - 7)^2
track <- nlm_track(f = himmelblau, p = c(0, 0))
summary(track)
#> Iterations: 16
#> Function improvement: 170 -> 1.521e-07
#> Computation time: 0.07373 seconds
#> Initial parameter: 0, 0
#> Final parameter: 3, 2
ggplot2::autoplot(track)
The next example tracks optim while it minimizes a
quartic polynomial:
polynomial <- function(x) 5 * x^4 + 4 * x^3 + x^2 + 3 * x + 2
gradient <- function(x) 20 * x^3 + 12 * x^2 + 2 * x + 3
track <- optim_track(
f = polynomial,
p = 0,
gradient = gradient,
method = "BFGS"
)
print(track)
#> # A tibble: 9 × 6
#> iteration value step parameter hessian seconds
#> * <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 0 2 0 0 NA 0
#> 2 1 0.344 -1.66 -0.6 9.20 0.00830
#> 3 2 0.241 -0.103 -0.672 13.0 0.00156
#> 4 3 0.212 -0.0293 -0.712 15.3 0.00144
#> 5 4 0.207 -0.00524 -0.730 16.4 0.00144
#> 6 5 0.206 -0.000679 -0.736 16.8 0.00143
#> 7 6 0.206 -0.0000752 -0.738 17.0 0.00271
#> 8 7 0.206 -0.00000784 -0.739 17.0 0.00151
#> 9 8 0.206 -0.000000802 -0.739 17.0 0.00147
ggplot2::autoplot(track)
If you have questions, find a bug, or need a feature, please file an issue on GitHub.