It sounds like you are hoping for an extension of latin hypercube designs to
non-rectangular regions. If such a thing exists, I'm not aware of it.
Would it be satisfactory to discard the points that violate the constraints?
You might have to experiment with different sample sizes to get the desired
number of acceptable points.
-- Tom
"Maninder " <m.khurana09@gmail.com> wrote in message
news:kg7cpf$1kk$1@newscl01ah.mathworks.com...
> Hi James,
> I wanted to use lhsdesign for producing parameter space within bounds and
> having certain constraints. I understand how to generate within given
> bounds through your reply but I am not able to impose the constraints on
> lhsdesign and the documentation of lhsdesign says nothing about
> constraints.
> I would really appreciate if you or someone else might answer my question.
> Thanks a lot.
>
>
> James Allison <james.allison@mathworks.com> wrote in message
> <hql1fr$red$1@fred.mathworks.com>...
>> You can always rescale lhsdesign output from [0 1] to another range. For
>> example, if lb and ub specify the lower and upper bounds for each
>> parameter, try:
>>
>> lb = [-10 0 0 10 100 0]; % lower bound
>> ub = [10 1 10 20 500 5]; % upper bound
>> n = 4; % number of samples
>> p = 6; % number of parameters
>> xn = lhsdesign(n,p); % generate normalized design
>> x = bsxfun(@plus,lb,bsxfun(@times,xn,(ub-lb)))
>>
>> -James
>>
>>
>> Md. Shahriar Karim wrote:
>> > Can anybody help me in generating a latin hypercube parameter space?
>> > I plan to use 6 variables with 4 different values of each. Is there any
>> > matlab function that can do all the tricks?
>> >
>> > I saw "lhsdesign (n, p)", but not sure how to utilize that for my
>> > parameter range. Thanks,
non-rectangular regions. If such a thing exists, I'm not aware of it.
Would it be satisfactory to discard the points that violate the constraints?
You might have to experiment with different sample sizes to get the desired
number of acceptable points.
-- Tom
"Maninder " <m.khurana09@gmail.com> wrote in message
news:kg7cpf$1kk$1@newscl01ah.mathworks.com...
> Hi James,
> I wanted to use lhsdesign for producing parameter space within bounds and
> having certain constraints. I understand how to generate within given
> bounds through your reply but I am not able to impose the constraints on
> lhsdesign and the documentation of lhsdesign says nothing about
> constraints.
> I would really appreciate if you or someone else might answer my question.
> Thanks a lot.
>
>
> James Allison <james.allison@mathworks.com> wrote in message
> <hql1fr$red$1@fred.mathworks.com>...
>> You can always rescale lhsdesign output from [0 1] to another range. For
>> example, if lb and ub specify the lower and upper bounds for each
>> parameter, try:
>>
>> lb = [-10 0 0 10 100 0]; % lower bound
>> ub = [10 1 10 20 500 5]; % upper bound
>> n = 4; % number of samples
>> p = 6; % number of parameters
>> xn = lhsdesign(n,p); % generate normalized design
>> x = bsxfun(@plus,lb,bsxfun(@times,xn,(ub-lb)))
>>
>> -James
>>
>>
>> Md. Shahriar Karim wrote:
>> > Can anybody help me in generating a latin hypercube parameter space?
>> > I plan to use 6 variables with 4 different values of each. Is there any
>> > matlab function that can do all the tricks?
>> >
>> > I saw "lhsdesign (n, p)", but not sure how to utilize that for my
>> > parameter range. Thanks,