compute.slices2: Compute the objective function on a grid of params and show...

In tlamadon/mopt: Parallel Derivative-free Moment Estimation

Description

Essentially the same as compute.slices, but does not report simulated moments but other model output. Useful for model output that is multidimensional. It's a simplified version of compute.slices in that it does not further process the model output: it return a list with nests "parameter name", "value of parameter", "model output". For example instead of reporting the mean of a certain statistic, this function can return a matrix or a higher dimensional array. Say you want to return the life-cycle profile of a certain model variable x. This will be a vector of length N, where N is the number of periods in the model. The user has to design the MOPT_OBJ_FUN in such a way that it returns the required output. There are 2 requirements for what MOPT_OBJ_FUN has to return. First it has to be a list, second, the list needs components "status" (indicating whether a particular evaluation is valid in some sense) and "output", which contains your custom model output.

Usage

compute.slices2(mcf, ns = 30, pad = 0.1, file = "est.slices.RData")

Arguments

mcf	object of class mopt
ns	number of points in each dimension to evaluate
pad	from bounds of parameter ranges. e.g. p in [0,1], avoid 0 and 1 with pad>0.
file	/path/to/your/file.RData

Value

list by paramter name, parameter value index, containing the value of the parameter vector and a list data containing your custom model output.

Examples

# generate custom output for a model
require(mopt)

# let's take a dummy objective function
MOPT_OBJ_FUNC <- objfc.norm2.custom(c(0,0),ns=2000)

# starting parameters
p <- list(x1=0.5,x2=0.5)
MOPT_OBJ_FUNC(p)

# then we want to setup the mopt
mcf                  = mopt_config(p)
mcf$wd               = getwd()
mcf$params_to_sample = c('x1','x2')
mcf$mode             = 'multicore'
mcf$algo             = algo.bgp


# set the parameter bounds
mcf <- mcf + 
  samplep('x1',-1,1) +
  samplep('x2',-1,1)


# prepare to run with OpenMP
require(parallel)
options(mc.cores = detectCores())

# finalize the preparation
mcf <- prepare.mopt_config(mcf)

# compute slices and generate plots
res <- compute.slices2(mcf,ns=30,pad=0.1)

tlamadon/mopt documentation built on May 31, 2019, 3:48 p.m.