logfill: Fill Logger

In martinzaefferer/COBBS: Continuous Optimization Benchmarks By Simulation

Fill Logger

Description

Fills logger result by last observation carried forward, to ensure that all runs (for each seed) have the same number of observations.

Usage

logfill(logs, freq)

Arguments

logs	the data frame produced by repeated experiments with, using the logger wrapper
freq	freq parameter of the logger

Value

a vector of scaled values

Examples

## first experiment to be logged
lower <- -1
upper <- 3
fun <- function(x)sqrt(abs(x-1))
require(nloptr)
lbfgs(x0=-1,
      fn = logger(fun,"mytest",1),
      lower=lower,
      upper=upper)
## store result
log <- as.data.frame(get(paste("logger","y","mytest",sep=".")))
colnames(log) <- c("it","y","cy")
log$seed <- 1 
## Repeat with different initial guess
lower <- -1
upper <- 3
fun <- function(x)sqrt(abs(x-1))
lbfgs(x0=0,
      fn = logger(fun,"mytest",1),
      lower=lower,
      upper=upper)
## store result
log2 <- as.data.frame(get(paste("logger","y","mytest",sep=".")))
colnames(log2) <- c("it","y","cy")
log2$seed <- 2
## combine both results
log <- rbind(log,log2)
plot(log$it) # second run was longer
## fill
logf <- logfill(log,1)
plot(logf$it) # filled in values for first run

martinzaefferer/COBBS documentation built on July 19, 2023, 4:12 a.m.