autosolve: autosolve

In amaendle/mvTargetOpt: Multivariate multi-target optimization

Description

Applies the iterative optimization algorithm suggested in this package. To perform the simulations automatically, the true model has to be specified. When in practice the true model is unknown, use opt.onestep function instead to get a candidate for the unknown optimum in a single iteration.

Usage

autosolve(
  startx,
  tgmean,
  tgerr,
  reps = 25,
  maxit = 10,
  reality = foo,
  xeps = 0.01,
  pplot = FALSE,
  pcnr = c(1, 2),
  maxarea = NULL,
  useweights = TRUE,
  mknormweights = F,
  gr2retlimit = T,
  mindeg = 0,
  sequential = F,
  tgpcawg = 1,
  yweights = F,
  datlim = NULL,
  knearest = NULL,
  tgdim = 1,
  ylast = NULL,
  sto = T,
  mod.sd = NULL,
  ...
)

Arguments

startx	numeric matrix, start values
tgmean	numeric v4ector, target value vector
tgerr	numeric vector, defines acceptable error range
reps	integer, number of repeated measurements
maxit	integer, maximum number of iterations
reality	function, real model
xeps	nunmeric, smallest (reasonably) distinguishable epsilon
pplot	boolean, diagnostic plots
pcnr	integer vector, defines which principal directions will be considered
maxarea	numeric matrix, area range, which will be explored
useweights	boolean
mknormweights	boolean
gr2retlimit	boolean
mindeg	integer, minimal degree of order of polynomial model
sequential	boolean
tgpcawg	numeric
yweights	boolean
datlim	NULL or integer
knearest	integer
tgdim	integer
ylast	integer
sto	boolean
mod.sd	numeric
...

Value

data.frame

Examples

# example 1: 2x2 MModell
tfoo <- function(x) {
  x1 <- x[,1]
  x2 <- x[,2]
  return( data.frame( y1=0.8*x1 - 1.2*x2,
                      y2=0.8*x1 - 1.2*abs(x2)^0.25
 ) ) }
dstgmean <- tfoo(cbind(1.35,1.4))
tgerr <- c(0.025,0.025)
xeps <- 0.01
startx <- expand.grid(x1=c(-1,3),x2=c(-1,3))
set.seed(123)
autosolve(startx,dstgmean,tgerr,reps=7,maxit=10,tfoo, xeps, F, pcnr=c(1,2), mod.sd=0.2)

# example 2, 3x3 model
set.seed(123)
startx <- data.frame(x1=runif(4,-4,4),x2=runif(4,-4,4),x3=runif(4,-4,4))
tfoo <- function(x) {
  x1 <- x[,1]
  x2 <- x[,2]
  x3 <- x[,3]
  return( data.frame( y1=0.8*x1 - 1.2*x2,
                      y2=0.8*x1 - 1.2*abs(x2)^0.25,
                      y3=0.8*x1 - 0.6*abs(x2)^0.25 + x3
 ) ) }
tgmean <- tfoo(cbind(1.35,1.4,1.5))#tfoo(cbind(0.35,0.4,0.5))
tgerr <- c(0.5,0.5,0.5)
tmp<-autosolve(startx,tgmean,tgerr*0.125,reps=4,maxit=6,tfoo, xeps=0.01, F, pcnr=c(1,2), mod.sd=0.2)

amaendle/mvTargetOpt documentation built on June 12, 2020, 5:57 p.m.