getConvergenceTimes: Get convergence times
In hdbeukel/james-analysis-R: Analysis Tools for the 'JAMES' Framework

Description Usage Arguments Details Value

Get the convergence times of the different runs of the given search applied to the given problem (in milliseconds). This is a generic S3 method.

1	getConvergenceTimes(data, problem, search, r = 0.99)

`data`	data object containing the analysis results
`problem`	name of the analyzed problem. Can be omitted if the `data` contains results for a single problem only.
`search`	name of the applied search. Can be omitted if the `data` contains results for a single search only (for the considered `problem`).
`r`	convergence ratio. Defaults to 0.99. Numeric value in [0,1].

If the data contains results for a single problem only, the argument problem can be omitted. Likewise, if for the considered problem results are available for a single search only, the argument search can be omitted.

The convergence time of a search run is defined as the time at which a certain value threshold is crossed. This threshold is computed from the given convergence ratio r as follows: if values are being maximized, thr = (1-r)*min + r*max; else, thr = (1-r)*max + r*min, where min and max are the minimum and maximum observed value in the considered search run, respectively. In case of maximization, a search run is said to have converged as soon as it reaches a value which is larger than or equal to the threshold thr. In case of minimization, convergence occurs when the values drop below the threshold. The convergence ratio r defaults to 0.99. If set to 1, a search run is said to have converged when the final best solution is found.

Numeric vector containing the convergence times of each run (in milliseconds). All convergence times are greater than or equal to -1.

hdbeukel/james-analysis-R documentation built on May 17, 2019, 3:18 p.m.