Nuclear criticality safety

NKO participants: C. Chevalier, Y. Richet (IRSN)

Unconstrained single objective optimization
2 variables, 75*75 data points, 40 noise levels

Video game data

NKO participant: M. Preuss

Test case A

22D highly noisy data from the car setup optimization competition setting (see manual concerning used programs and list of parameters), performed with the standard car and bot on track wheel-2 (Suzuka F1).

The objective function is the -1 * distance travelled in 2000 tics (40 s) divided by 2000 if the damage is at most 100, and the damage else (values up to 10000) according to recommendations in the diploma thesis of Markus Kemmerling (in German), see also this paper: Automatic Adaptation to Generated Content Via Car Setup Optimization in TORCS contains the following files:

inputs125p.txt Sample locations 125 points for 4×125, LHS-distributed)
inputs250p.txt Sample locations 250 points for 2×250, LHS-distributed)
inputs500p.txt Sample locations 500 points for 1×500, LHS-distributed)
results-125p-1-4.txt Results for inputs125p.txt, in a row (1-125,1-125,1-125,1-125)
results-250p-1-2.txt Results for inputs125p.txt, in a row (1-250,1-250)
results-500p.txt Results for inputs500p.txt
newTorcsValidation2.txt Validation data (440 valid points, LHS-distributed)

Note that if the evaluation of a car was not finished due to damage (exceeding 10000), the result value is NA. In the validation set, we removed all cars where this happened.

Hydrogeologic Testcase: well placement

NKO participant: C. Bürger

To be found: Optimal drinking water well locations (xi and yi coordinates) and extraction rate(Qi) for a number of wells, ntot, with i = 1,…ntot.

The optimization problem dimension is variable: 3*ntot (ntot between one and five). The noise comes from uncertainties on the underground hydraulic conductivity. The problem has boundary constraints and three non-trivial constraints, which can be taken into account with a penalization technique.

Technical details:

- the application runs in Matlab and calls a couple of batch scripts and Windows executables

- a detailed notice is provided

- an example of matlab function for optimization with cma-es is provided

- all the files are downloadable here:

Plant irigation problem

NKO participant: M. de Paly

1000 realizations, 15 to 100 variables (irrigation times and amounts)

BBOB test case

(C++ / Java with matlab interface)

SPO test case

NKO participant: M. Preuss

Financial test case

NKO participant: Didier Rullière

Last modified: 2015-09-08 15:53 (external edit)