historize: Historize function

In adagio: Discrete and Global Optimization Routines

Historize function

Description

Provides storage for function calls.

Usage

Historize(fun, len = 0, ...)

Arguments

fun	Function of one or several variables.
len	If > 0, input values will be stored, too.
...	Additional parameters to be handed to the function.

Details

Historize() adds storage to the function. If len=0 then only function values will be stored in a vector, if len>0 then variables will be stored in a vector, and function values will be stored in a matrix, one line for every function call.

If Fn = Historize(...) is the 'historized' function, then the storage can be retrieved with Fn(), and can be cleared with Fn(NULL).

Filling the storage will take extra time and can slow down the function calls. Especially also storing the variables used in the call (with len>0) will make it considerably slower.

Functions can have multivariate output; the user is asked to take care of handling the output vector or matrix correctly. The function may even require additional parameters.

Value

Returns a list, $input the input variables as matrix, $H2 the function values as vector, $nvars the number of input variables of the function, and $ncalls the number or recorded function calls.

Note

Can also be applied to functions that output a vector (same length for every call).

Author(s)

Hans W. Borchers

See Also

trace

Examples

f <- function(x) sum(x^2)
F <- Historize(f, len = 1)
c( F(c(1,1)), F(c(1,2)), F(c(2,1)), F(c(2,2)) )
#> [1] 2 5 5 8
F()
#> $input
#>      [,1] [,2]
#> [1,]    1    1
#> [2,]    1    2
#> [3,]    2    1
#> [4,]    2    2
#> 
#> $values
#> [1] 2 5 5 8
#> 
#> $nvars
#> [1] 2
#> 
#> $ncalls
#> [1] 4
F(NULL)     # reset memory


## Rastrigin under Differential Evolution

Fn <- Historize(fnRastrigin)
dm <- 10
lb <- rep(-5.2, dm); ub <- -lb

sol <- simpleDE(Fn, lower = lb, upper = ub)
fvalues <- Fn()$values
fvals <- cummin(fvalues)

## Not run: 
plot(fvalues, type = 'p', col = 7, pch = '.', cex = 2,
     main = "Simple DE optimization", xlab = '', ylab = '')
lines(fvals, col = 2, lwd = 2)
legend(length(fvalues), max(fvalues),
       c("Intermediate values", "cummulated min"),
       xjust = 1, col = c(7, 2), lwd = 2)
grid()
## End(Not run)

adagio documentation built on Oct. 26, 2023, 5:08 p.m.