UniDOE (version 1.0)

UDC: Uniform Design Construction

Description

This function takes n,s,q and other arguments to output a list(described below).

Usage

UDC(n,s,q,init,initX,crit,maxiter,hits_ratio,vis)

Arguments

n

an integer R object. Run of Experiment

s

an integer R object. Factor of Experiment.

q

an integer R object. Level of Experiment.

init

an string vector R object: "rand"(default), "orth" or "input". Criterion used

initX

an user-defined integer matrix R object, which is fixed in search. Needed when init="orth" or init="input"

crit

an character R object. Type of criterion to use.

"maximin" -- maximin Discrepancy ;

"CL2" --Centered L2 Discrepancy ;

"MD2" --Mixture L2 Discrepancy ;

maxiter

a positive integer R object. Maximum iteration number in outer while loop of SOAT algorithm.

hits_ratio

an float R object. Default value is 0.1, which is the ratio to accept changes of design in inner for loop. Details can be checked in (Zhang, A. and Li, H. (2017). UniDOE: an R package for uniform design construction via stochastic optimization.)

vis

an boolean R object. If true, plot the criterion value sequence.

Value

A list that contains Initial design matrix(initial_design),optimal design matrix(final_design), initial criterion value(initial_criterion), final criterion value(criterion_value) and criterion list(criterion_lists) in update process.

References

Zhang, A. and Li, H. (2017). UniDOE: An R package for constructing uniform design of experiments via stochastic and adaptive threshold accepting algorithm. Technical Report.

Examples

# NOT RUN {
##e.g.1
n=12#(must be multiples of q)
s=3
q=4
crit = "MD2"#(Mixture L2 criteria)
list1=UDC(n,s,q,crit=crit,maxiter=100)

##e.g.2
n=10 #(must be multiples of q)
s=3
q=5
init = "rand"
crit = "MD2" #(Mixture L2 criteria)
vis=TRUE
list1=UDC(n,s,q,init=init,crit=crit,maxiter=100,vis=vis)

##e.g.3
#If init = "orth",
#algorithm will fix columns of initX and expand total number of columns to s for final design.
n=6
s=5
q=3
init = "orth"
# initX should be orthogonal design with nrow=n&level=q
initX = matrix(c(1,1,2,2,3,3,3,3,1,1,2,2),ncol=2)
list = UDC(n,s,q,init=init,initX = initX,maxiter=100)

##e.g.4
#If init="input", algorithm will search for better a better design with same size as initX.
init = "input"
# initX should be orthogonal design with nrow=n&level=q
initX = matrix(c(1,1,2,2,3,3,3,3,1,1,2,2),ncol=2)
list = UDC(init=init,initX = initX,maxiter=100)

# }