MGH10_opt: More, Gabrow and Hillstrom example 10

In NISTopt: Nonlinear least squares examples from NIST in form of functions for optim() and optimx()

Description

The MGH10 data frame has 16 rows and 2 columns.

Format

This data frame contains the following columns:

y

A numeric vector of response values.

x

A numeric vector of input values.

Details

This problem was found to be difficult for some very good algorithms.

See More, J. J., Garbow, B. S., and Hillstrom, K. E. (1981). Testing unconstrained optimization software. ACM Transactions on Mathematical Software. 7(1): pp. 17-41.

Source

Meyer, R. R. (1970). Theoretical and computational aspects of nonlinear regression. In Nonlinear Programming, Rosen, Mangasarian and Ritter (Eds). New York, NY: Academic Press, pp. 465-486.

Examples

Try <- function(expr) if (!inherits(val <- try(expr), "try-error")) val
plot(y ~ x, data = MGH10)
## check plot on log scale for shape
plot(y ~ x, data = MGH10, log = "y")
## starting values for this run are ridiculous
Try(fm1 <- nls(y ~ b1 * exp(b2/(x+b3)), data = MGH10, trace = TRUE,
           start = c(b1 = 2, b2 = 400000, b3 = 25000)))
Try(fm1a <- nls(y ~ b1 * exp(b2/(x+b3)), data = MGH10,
                trace = TRUE, alg = "port",
                start = c(b1 = 2, b2 = 400000, b3 = 25000)))
Try(fm2 <- nls(y ~ b1 * exp(b2/(x+b3)), data = MGH10, trace = TRUE,
           start = c(b1 = 0.02, b2 = 4000, b3 = 250)))
Try(fm2a <- nls(y ~ b1 * exp(b2/(x+b3)), data = MGH10,
                trace = TRUE, alg = "port",
                start = c(b1 = 0.02, b2 = 4000, b3 = 250)))
Try(fm3 <- nls(y ~ exp(b2/(x+b3)), data = MGH10, trace = TRUE,
               start = c(b2 = 400000, b3 = 25000),
               algorithm = "plinear"))
Try(fm4 <- nls(y ~ exp(b2/(x+b3)), data = MGH10, trace = TRUE,
           start = c(b2 = 4000, b3 = 250),
           algorithm = "plinear"))

NISTopt documentation built on July 25, 2017, 3:01 p.m.