calibrar
packageThis package allows the parameter estimation (i.e. calibration) of
complex models, including stochastic ones. It implements generic
functions that can be used for fitting any type of models, especially
those with non-differentiable objective functions, with the same syntax
as base::optim
. It supports multiple phases estimation
(sequential parameter masking), constrained optimization (bounding box
restrictions) and automatic parallel computation of numerical gradients.
Some common maximum likelihood estimation methods and automated
construction of the objective function from simulated model outputs is
provided.
This vignette covers the basic usage of the package, introducing the
functions optim2()
, optimh()
and
calibrate()
.
As the name sugests, optim2()
is intended to extend the
functionality of stats::optim()
and it uses the same
arguments (with some additions):
optim2(
par,
fn,
gr = NULL,
...,
method = c("Nelder-Mead", "BFGS", "CG", "L-BFGS-B", "SANN", "Brent", "nlm", "nlminb",
"Rcgmin", "Rvmmin", "hjn", "spg", "LBFGSB3", "AHR-ES"),
lower = -Inf,
upper = +Inf,
active = NULL,
control = list(),
hessian = FALSE,
parallel = FALSE
)
The first difference is the possible values for the
method
argument. In addition to the first six methods, also
available in optim()
, optim2()
gives access to
stats::nlm()
and stats::nlminb()
but with the
same syntax as optim()
to make them easy to use. In
addition, three methods from the optimr
package
(Rcgmin
, Rvmmin
, hjn
), the
L-BFGS-B v3 implemented in the bfgsb3c
package and the
AHR-ES
(Adaptative Hierarchical Recombination Evolutionary
Strategy) implemented in this package.
In the next example, we compare the outputs of optim()
and optim2()
:
library(calibrar)
optim(par=rep(1, 5), fn=function(x) sum(x^2))
#> $par
#> [1] -1.931714e-04 -3.044063e-04 -1.744066e-04 1.940552e-05 5.641574e-05
#>
#> $value
#> [1] 1.639553e-07
#>
#> $counts
#> function gradient
#> 456 NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> NULL
optim2(par=rep(1, 5), fn=function(x) sum(x^2))
#> $par
#> [1] -1.931714e-04 -3.044063e-04 -1.744066e-04 1.940552e-05 5.641574e-05
#>
#> $value
#> [1] 1.639553e-07
#>
#> $counts
#> function gradient
#> 456 NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> NULL
The results are identical, as here optim2()
acts just as
a wrapper for optim()
. Now, we can run the same example
with two other methods:
optim2(par=rep(1, 5), fn=function(x) sum(x^2), method="nlm")
#> $par
#> [1] -2.502443e-13 -2.502443e-13 -2.502443e-13 -2.502443e-13 -2.502443e-13
#>
#> $value
#> [1] 3.13111e-25
#>
#> $counts
#> function gradient
#> NA NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> [1] "Relative gradient is close to zero, current iterate is probably solution."
set.seed(880820) # for reproducibility
optim2(par=rep(1, 5), fn=function(x) sum(x^2), method="AHR-ES")
#> $par
#> [1] 3.749723e-10 -2.023475e-10 1.655858e-10 -2.272363e-10 1.488461e-10
#>
#> $value
#> [1] 2.827589e-19
#>
#> $counts
#> function gradient
#> 1024 0
#>
#> $convergence
#> [1] 0
#>
#> $message
#> [1] "Stopping criteria reached in 128 generations."
#>
#> $generations
#> [1] 128
The second difference is the new argument active
, which
is a vector indicating if a parameters will be optimized (i.e. active)
or fixed to a constant value during the optimization process. In the
next example, we will fix the third and fourth parameters to its initial
values:
optim2(par=rep(1, 5), fn=function(x) sum(x^2),
active=c(TRUE, TRUE, FALSE, FALSE, TRUE))
#> Warning in paropt[isActive] <- output$par: número de items para para sustituir
#> no es un múltiplo de la longitud del reemplazo
#> $par
#> [1] 1.062264e-05 1.167696e-04 1.000000e+00 1.000000e+00 5.492100e-05
#>
#> $value
#> [1] 2
#>
#> $counts
#> function gradient
#> 202 NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> NULL
As we can see, in the final solution, the par
value keep
the values at 1, the initial value provided. All the numerical gradients
computed internally have also ‘masked’ this parameters and the
derivatives are not computed for them to speed up computation time.
Finally, the third difference is the new argument
parallel
, that active the parallel computation of the
numerical gradient, when gr
is not supplied:
optim2(par=rep(1, 5), fn=function(x) sum(x^2), parallel=TRUE)
#> $par
#> [1] -1.931714e-04 -3.044063e-04 -1.744066e-04 1.940552e-05 5.641574e-05
#>
#> $value
#> [1] 1.639553e-07
#>
#> $counts
#> function gradient
#> 456 NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> NULL
This last option will increase performance when the computation time
of fn
in considerable. We will explain in detail this
feature in a following section.
Additionally, the method for the computation of the numerical
gradient can be chosen within the control
list:
The function optimh()
has a similar functionality as
optim2()
but acts as a wrapper for several heuristic
optimization algorithms implemented in several packages:
dfoptim
, optimr
, minqa
,
cmaes
, genSA
, DEoptim
,
soma
, rgenoud
and psoptim
. The
optimh()
function standardizes the inputs and outputs to
those of optim()
, providing a more convenient user
interface. All specific arguments of this methods can be passed to the
original function using the control
argument.
optimh(
par,
fn,
gr = NULL,
...,
method = c("AHR-ES", "Nelder-Mead", "SANN", "hjn", "CMA-ES", "genSA", "DE", "soma",
"genoud", "PSO", "hybridPSO", "mads", "hjk", "hjkb", "nmk", "nmkb"),
lower = -Inf,
upper = +Inf,
active = NULL,
control = list(),
hessian = FALSE,
parallel = FALSE
)
# Covariance Matrix Adaptation Evolutionary Strategy
set.seed(880820) # for reproducibility
optimh(par=rep(1, 5), fn=function(x) sum(x^2), method="CMA-ES",
control=list(maxit=200))
#> $par
#> [1] -1.848860e-11 3.646458e-11 -8.209875e-11 2.866646e-11 -8.636439e-11
#>
#> $value
#> [1] 1.669227e-20
#>
#> $counts
#> function gradient
#> 1600 NA
#>
#> $convergence
#> [1] 1
#>
#> $message
#> NULL
# Generalized Simulated Anneling
set.seed(880820) # for reproducibility
optimh(par=rep(1, 5), fn=function(x) sum(x^2), method="genSA",
lower=rep(-100, 5), upper=rep(100, 5),
control=list(maxit=200, temperature=6000))
#> $par
#> [1] 6.988898e-11 6.988898e-11 6.988898e-11 6.988898e-11 6.988898e-11
#>
#> $value
#> [1] 2.442235e-20
#>
#> $counts
#> function gradient
#> 2024 0
#>
#> $convergence
#> [1] 0
#>
#> $message
#> [1] NA
# Self-Organising Migrating Algorithm
set.seed(880820) # for reproducibility
optimh(par=rep(1, 5), fn=function(x) sum(x^2), method="soma",
lower=rep(-100, 5), upper=rep(100, 5),
control=list(maxit=200))
#> $par
#> [1] 1.519459e-16 -4.704646e-16 8.741813e-17 1.169846e-17 -1.104563e-16
#>
#> $value
#> [1] 2.644038e-31
#>
#> $counts
#> function gradient
#> 2000 NA
#>
#> $convergence
#> [1] 0
#>
#> $message
#> [1] NA
The maxit
control argument has been standardized to work
with all methods and to represent the maximum number of iterations of
the algorithm. However, the specific number of function evaluations per
iteration may vary between methods. You can refer to the help pages from
each package to have details about every specific method and its control
arguments.
Most algorithms implemented in optim2()
and some in
optimh
can benefit of parallel computation. For the methods
that uses the numerical computation of the gradient, this will be
calculated in parallel. In order to support any type of parallel
implementation, the parallel setup is NOT automatic, and must be done by
the user previous to executed the optimization, as described in the
following example:
library(parallel)
ncores = detectCores() - 1 # number of cores to be used
cl = makeCluster(ncores)
# this is slower than sequential for very fast models (like this one)
optim2(par=rep(0.5, 5), fn=function(x) sum(x^2),
control=list(ncores=ncores), parallel=TRUE)
stopCluster(cl) # close the parallel connections
The calibrate()
function implements an automatic
sequential parameter estimation, meaning parameters can be un-masked
(set active) progressively during sequential phases of the calibration
process.
calibrate(
par,
fn,
gr = NULL,
...,
method = NULL,
lower = NULL,
upper = NULL,
phases = NULL,
control = list(),
hessian = FALSE,
replicates = 1,
parallel = FALSE
)
With the basic syntax, the calibrate()
function will
work similarly to optim2()
and optimh()
,
performing a simple optimization:
calibrate(par=c(1,2,3,NA,NA), fn=function(x) sum(x^2))
#> Using optimization method 'Rvmmin'.
#> Elapsed time: 0.00s
#> Function value: 1.24979e-16
#> Parameter values: -5e-09 -5e-09 -5e-09 -5e-09 -5e-09
#>
#> Status: Rvmminu appears to have converged
#> Optimization using 'Rvmmin' algorithm.
#> Function value: 1.249787e-16
#> Status: Rvmminu appears to have converged
#> Parameters:
#> [1] -4.999645e-09 -4.999291e-09 -4.998935e-09 -5.000000e-09 -5.000000e-09
#> Computation:
#> function gradient
#> 10 6
If upper
and lower
bounds are provided, the
calibrate function can take NA
as starting values for the
optimization, to some or all of the parameters:
calibrate(par=c(1,2,3,NA,5), fn=function(x) sum(x^2),
lower=rep(-100, 5), upper=rep(100, 5))
#> Using optimization method 'Rvmmin'.
#> Elapsed time: 0.00s
#> Function value: 1.23526e-16
#> Parameter values: -4.98e-09 -4.97e-09 -4.96e-09 -5e-09 -4.94e-09
#>
#> Status: Rvmminb appears to have converged
#> Optimization using 'Rvmmin' algorithm.
#> Function value: 1.235263e-16
#> Status: Rvmminb appears to have converged
#> Parameters:
#> [1] -4.982589e-09 -4.974351e-09 -4.960025e-09 -4.999959e-09 -4.935026e-09
#> Computation:
#> function gradient
#> 13 5
Multiple phases
can be set up by selecting a different
one for each parameter. The starting value of the optimization for each
phase is updated with the best parameters found in the previous
phase:
calibrate(par=c(1,2,3,NA,5), fn=function(x) sum(x^2),
lower=rep(-100, 5), upper=rep(100, 5),
phases=c(1,2,3,2,1))
#> Parameter estimation in three phases.
#>
#> - Phase 1: 2 of 5 parameters active.
#> Using optimization method 'Rvmmin'.
#> Phase 1 finished (0.00s)
#> Function value: 13
#> Parameter values: -2.97e-08 -2.61e-08
#>
#> - Phase 2: 4 of 5 parameters active.
#> Using optimization method 'Rvmmin'.
#> Phase 2 finished (0.00s)
#> Function value: 9
#> Parameter values: 5.84e-09 -3.93e-08 0 9.41e-09
#>
#> - Phase 3: 5 of 5 parameters active.
#> Using optimization method 'Rvmmin'.
#> Phase 3 finished (0.00s)
#> Function value: 1.24997e-16
#> Parameter values: -5e-09 -5e-09 -5e-09 -5e-09 -5e-09
#>
#> Status: Rvmminb appears to have converged
#> Optimization using 'Rvmmin' algorithm.
#> Function value: 1.24997e-16
#> Status: Rvmminb appears to have converged
#> Parameters:
#> [1] -5.000004e-09 -5.000078e-09 -4.999605e-09 -5.000002e-09 -5.000006e-09
#> Computation:
#> function gradient
#> 11 6
When a phase is set to a negative number, the parameter is fixed at its initial value during all the calibration and it is never optimized:
calibrate(par=c(1,2,3,NA,5), fn=function(x) sum(x^2),
lower=rep(-100, 5), upper=rep(100, 5),
phases=c(1,2,-1,2,1))
#> Parameter estimation in two phases.
#>
#> - Phase 1: 2 of 5 parameters active.
#> Using optimization method 'Rvmmin'.
#> Phase 1 finished (0.00s)
#> Function value: 13
#> Parameter values: -2.97e-08 -2.61e-08
#>
#> - Phase 2: 4 of 5 parameters active.
#> Using optimization method 'Rvmmin'.
#> Phase 2 finished (0.00s)
#> Function value: 9
#> Parameter values: 5.84e-09 -3.93e-08 0 9.41e-09
#>
#> Status: Rvmminb appears to have converged
#> Optimization using 'Rvmmin' algorithm.
#> Function value: 9
#> Status: Rvmminb appears to have converged
#> Parameters:
#> [1] 5.835174e-09 -3.934291e-08 3.000000e+00 0.000000e+00 9.410321e-09
#> * Some parameters are not calibrated.
#> Computation:
#> function gradient
#> 6 4
When dealing with stochastic functions, the argument
replicates
can be helpful, as it allows to evaluate the
objective function several times, taking the average value of them as
the actual function value (as an approximation of the expected value of
the function). When replicates=1
, the algorithm used by
default is “LBFGSB3”, but when replicates is greater than 1, “AHR-ES” is
used. The next examples use the function sphereN()
, which
computes the Euclidean distance from a point x
to the
origin of coordinates after a random displacement of its position (see
?sphereN
for details). We will set the maximum number of
iterations to 1000 to speed up the execution of the vignette.
calibrate(par=c(1,2,3,NA,5), fn=sphereN,
lower=rep(-100, 5), upper=rep(100, 5),
phases=c(1,2,3,2,1), replicates=3, control=list(maxit=1000))
#> Parameter estimation in three phases.
#>
#> - Phase 1: 2 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 1 finished (0.80s)
#> Function value: 12.9126
#> Parameter values: -0.324 0.12
#>
#> - Phase 2: 4 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 2 finished (0.90s)
#> Function value: 10.1169
#> Parameter values: 0.0695 0.112 0.151 -0.0849
#>
#> - Phase 3: 5 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 3 finished (0.91s)
#> Function value: 0.0534613
#> Parameter values: -0.00843 0.0767 0.0606 0.0265 0.0048
#>
#> Status: Maximum number of generations or function evaluations reached.
#> Optimization using 'AHR-ES' algorithm.
#> Function value: 0.05346132
#> Status: Maximum number of generations or function evaluations reached.
#> Parameters:
#> [1] -0.008425148 0.076711662 0.060580990 0.026532747 0.004796660
#> Computation: (1000 generations)
#> function gradient
#> 8000 0
The number of replicates can be one single value or a vector with the length equal to the number of phases:
calibrate(par=c(1,2,3,NA,5), fn=sphereN,
lower=rep(-100, 5), upper=rep(100, 5),
phases=c(1,2,3,2,1), replicates=c(1,1,5), control=list(maxit=1000))
#> Parameter estimation in three phases.
#>
#> - Phase 1: 2 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 1 finished (0.67s)
#> Function value: 13.3163
#> Parameter values: -0.346 -0.101
#>
#> - Phase 2: 4 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 2 finished (0.73s)
#> Function value: 9.8584
#> Parameter values: 0.217 -0.335 0.24 0.101
#>
#> - Phase 3: 5 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 3 finished (1.16s)
#> Function value: 0.0412229
#> Parameter values: 0.0282 0.00574 -0.0127 -0.0375 -0.0192
#>
#> Status: Maximum number of generations or function evaluations reached.
#> Optimization using 'AHR-ES' algorithm.
#> Function value: 0.04122286
#> Status: Maximum number of generations or function evaluations reached.
#> Parameters:
#> [1] 0.028236432 0.005735164 -0.012675352 -0.037462605 -0.019210042
#> Computation: (1000 generations)
#> function gradient
#> 8000 0
calibrate(par=list(par1=c(1,2,3), par2=NA, par3=5), fn=sphereN,
lower=rep(-100, 5), upper=rep(100, 5),
phases=c(1,2,-3,2,1), replicates=c(1,5), control=list(maxit=1000))
#> Parameter estimation in two phases.
#>
#> - Phase 1: 2 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 1 finished (0.90s)
#> Function value: 13.1182
#> Parameter values: 0.0558 0.181
#>
#> - Phase 2: 4 of 5 parameters active.
#> Using optimization method 'AHR-ES'.
#> Phase 2 finished (1.65s)
#> Function value: 8.96767
#> Parameter values: -0.0167 0.164 -0.14 0.0335
#>
#> Status: Maximum number of generations or function evaluations reached.
#> Optimization using 'AHR-ES' algorithm.
#> Function value: 8.96767
#> Status: Maximum number of generations or function evaluations reached.
#> Parameters:
#> par11 par12 par13 par2 par3
#> -0.01674863 0.16445070 3.00000000 -0.13951818 0.03351795
#> * Some parameters are not calibrated.
#> Computation: (1000 generations)
#> function gradient
#> 8000 0
Note that the function fn
must be able to take a list as
parameter set, and the user must ensure this works beforehand.
Most algorithms implemented in the calibrate
function
can benefit of parallel computation. For the methods that uses the
numerical computation of the gradient, this will be calculated in
parallel as in optim2()
. In order to support any type of
parallel implementation, the parallel setup is NOT automatic, and must
be done by the user previous to executed the optimization, as described
in the following example:
library(parallel)
ncores = detectCores() - 1 # number of cores to be used
cl = makeCluster(ncores)
# this is slower than sequential for very fast models (like this one)
calib = calibrate(par=rep(0.5, 5), fn=sphereN,
replicates=3,
lower=rep(-5, 5),
upper=rep(+5, 5),
phases=c(1,1,1,2,3),
control=list(parallel=TRUE, ncores=ncores))
stopCluster(cl) # close the parallel connections
Please, refer to vignette(package="calibrar")
for
additional vignettes or to the calibrar website
for more details.