optimx (version 2018-7.10)

scalechk: Check the scale of the initial parameters and bounds input to an optimization code used in nonlinear optimization

Description

Nonlinear optimization problems often have different scale for different parameters. This function is intended to explore the differences in scale. It is, however, an imperfect and heuristic tool, and could be improved.

At this time scalechk does NOT take account of masks. (?? should 110702)

Usage

scalechk(par, lower = lower, upper = upper, bdmsk=NULL, dowarn = TRUE)

Arguments

par

A numeric vector of starting values of the optimization function parameters.

lower

A vector of lower bounds on the parameters.

upper

A vector of upper bounds on the parameters.

bdmsk

An indicator vector, having 1 for each parameter that is "free" or unconstrained, and 0 for any parameter that is fixed or MASKED for the duration of the optimization.

dowarn

Set TRUE to issue warnings. Othwerwise this is a silent routine. Default TRUE.

Value

A list with components:

# Returns: # list(lpratio, lbratio) -- the log of the ratio of largest to smallest parameters # and bounds intervals (upper-lower) in absolute value (ignoring Inf, NULL, NA)

lpratio

The log of the ratio of largest to smallest parameters in absolute value (ignoring Inf, NULL, NA)

lbration

The log of the ratio of largest to smallest bounds intervals (upper-lower) in absolute value (ignoring Inf, NULL, NA)

Details

The scalechk function will check that the bounds exist and are admissible, that is, that there are no lower bounds that exceed upper bounds.

NOTE: Free paramters outside bounds are adjusted to the nearest bound. We then set parchanged = TRUE which implies the original parameters were infeasible.

There is a check if lower and upper bounds are very close together, in which case a mask is imposed and maskadded is set TRUE. NOTE: it is generally a VERY BAD IDEA to have bounds close together in optimization, but here we use a tolerance based on the double precision machine epsilon. Thus it is not a good idea to rely on scalechk() to test if bounds constraints are well-posed.

Examples

# NOT RUN {
#####################
  par <- c(-1.2, 1)
  lower <- c(-2, 0)
  upper <- c(100000, 10)
  srat<-scalechk(par, lower, upper,dowarn=TRUE)
  print(srat)
  sratv<-c(srat$lpratio, srat$lbratio)
  if (max(sratv,na.rm=TRUE) > 3) { # scaletol from ctrldefault in optimx
     warnstr<-"Parameters or bounds appear to have different scalings.\n
     This can cause poor performance in optimization. \n
     It is important for derivative free methods like BOBYQA, UOBYQA, NEWUOA."
     cat(warnstr,"\n")
  }

# }

Run the code above in your browser using DataCamp Workspace