james: Results of example algorithm comparison

In james.analysis: Analysis Tools for the 'JAMES' Framework

Description

Contains results of an example analysis performed with the 'JAMES' extensions module. The performance of two algorithms is compared (random descent and parallel tempering) for a core selection problem in which the mean entry-to-nearest-entry distance is maximized. Four different data sets have been analyzed. Details about the performed analysis are provided at the website (see below).

Usage

james

Format

S3 object of class "james", as if produced by readJAMES.

Source

http://www.jamesframework.org/examples/#analysis

See Also

readJAMES

Examples

# load data
data(james)
summary(james)

# plot convergence curves for coconut data set
plotConvergence(james, problem = "coconut", min.time = 1000, max.time = 100000)

# create box plots of solution values (quality) and convergence times
boxplot(james, problem = "coconut")
boxplot(james, problem = "coconut", type = "time")

# extract solution values and convergence times for parallel tempering and random descent
values.pt <- getBestSolutionValues(james, problem = "coconut", search = "Parallel Tempering")
times.pt <- getConvergenceTimes(james, problem = "coconut", search = "Parallel Tempering")
values.rd <- getBestSolutionValues(james, problem = "coconut", search = "Random Descent")
times.rd <- getConvergenceTimes(james, problem = "coconut", search = "Random Descent")

# perform wilcoxon test to compare distributions across algorithms
values.test <- wilcox.test(values.pt, values.rd)
values.test
times.test <- wilcox.test(times.pt, times.rd)
times.test

# adjust p-values for multiple testing
p.adjust(c(values.test$p.value, times.test$p.value))

Example output

Problem:         Search:             Runs:  Mean value:  St. dev:   Median:      IQR:
---------------  ------------------  -----  -----------  --------  --------  --------
coconut          Parallel Tempering     10        0.571  7.15e-05     0.571  0.000113
coconut          Random Descent         10         0.57  0.000553      0.57  0.000754
maize-accession  Parallel Tempering     10        0.579         0     0.579         0
maize-accession  Random Descent         10        0.576  0.000673     0.576   0.00106
maize-bulk       Parallel Tempering     10         0.43         0      0.43         0
maize-bulk       Random Descent         10        0.429  0.000526     0.429  0.000394
pea-small        Parallel Tempering     10        0.354  0.000274     0.354  0.000375
pea-small        Random Descent         10        0.351  0.000477     0.351  0.000545

	Wilcoxon rank sum test

data:  values.pt and values.rd
W = 96, p-value = 0.0001299
alternative hypothesis: true location shift is not equal to 0


	Wilcoxon rank sum test

data:  times.pt and times.rd
W = 78, p-value = 0.03546
alternative hypothesis: true location shift is not equal to 0

[1] 0.0002598021 0.0354629890

james.analysis documentation built on May 1, 2019, 9:13 p.m.