[R] constrOptim with method = "L-BFGS-B"

lhaba georgeshabre at gmail.com
Wed Jun 18 09:31:48 CEST 2008


Thank you for your answer. 

I posted this problem as it is because I am benchmarking multiple solvers
over this particular problem. 

I will enquire LowRankQP, kernlab and quadprog packages as you suggested.

Thank you

Georges


Spencer Graves wrote:
> 
>       I believe that 'optim' will not accept equality constraints. 
> 
>       However, you do not need the generality of 'optim' to "minimize a 
> quadratic function with boundary conditions and one equality 
> condition".  This type of problem is called "quadratic programming", 
> and  RSiteSearch("quadratic programming", "fun") just returned 29 hits 
> for me.  The first 3 cite functions in the "LowRankQP", "kernlab", and 
> "quadprog" packages.  I don't know if any of these will solve your 
> problem, but I suspect that at least one might.  If not, can you recast 
> the problem to remove the equality constraint? 
> 
>       If the above does not work for you, I suggest you try a much 
> simpler version, e.g., with 'mat' = a 3 x 3 array with one inequality 
> and one equality, as suggested in the famous book by Polya on "How to 
> Solve It" (http://en.wikipedia.org/wiki/How_to_Solve_It).  This has the 
> added advantage of giving you a simpler example to send to this list if 
> you can't make it work.  You are to be commended for providing a 
> self-contained example.  Unfortunately, your example is so large that it 
> is slightly intimidating.  A simpler example might elicit more (and more 
> useful) replies -- if it doesn't lead you to the solution, as Polya 
> suggested that it might. 
> 
>       Hope this helps,
>       Spencer     
> 
> lhaba wrote:
>> Hi, 
>> i need to minimize a quadratic function with boundary condidtions and one
>> equality condition. 
>> In order to do that i converted the equality constraint into 2 inequality
>> constaints and passed everything cia constrOptim, as the manual said:
>> everything included in the ... will be passed to Optim that will pass it
>> back to fn in case it does not need it.
>>
>> My code is the following:
>>
>> mat <- array( c(0.0001088799073581,	0.0000136029502036,
>> 0.0000060430384243,
>> 0.0000847097879033,	0.0000115053365822,	0.0000216245975292,
>> 0.0000483253391811,	0.0000787580901352,	0.0000186474817658,
>> 0.0000312260571354,	0.0000217594093734,	0.0000536298150897,
>> 0.0000166202592455,	0.0000451975061637,	-0.0000120364862228,
>> 0.0000497117714376,
>> 0.0000136029502036,	0.0001537319301276,	0.0000226518408080,
>> 0.0000591480002102,	0.0000797128619950,	0.0000091332643423,
>> 0.0000693354260457,	0.0000825217915015,	0.0000229122227269,
>> 0.0000297662414650,	0.0000334443258658,	0.0000273254534933,
>> 0.0000202062301763,	0.0000026260702295,	0.0000558975248740,
>> 0.0000953647537111,
>> 0.0000060430384243,	0.0000226518408080,	0.0005971325756834,
>> -0.0000762583321100,	-0.0000246005202071,	-0.0000300982253054,
>> 0.0000299178429478,	0.0000135672602503,	0.0001735431064391,
>> -0.0000133347388414,	0.0001387582890571,	0.0000964898243724,
>> -0.0000149571346672,	0.0000104437939143,	0.0001246900353191,
>> -0.0000171884354549,
>> 0.0000847097879033,	0.0000591480002102,	-0.0000762583321100,
>> 0.0004968467836203,	0.0002303499425964,	0.0000992731601466,
>> 0.0002685466918035,	0.0002580180069951,	0.0000725833959653,
>> 0.0000525639940758,	0.0001785049461665,	0.0001781339191317,
>> 0.0000597631329497,	0.0000201160486244,	0.0002582267884874,
>> 0.0002473268250781,
>> 0.0000115053365822,	0.0000797128619950,	-0.0000246005202071,
>> 0.0002303499425964,	0.0002945009393242,	-0.0000426583313588,
>> 0.0002067711081561,	0.0002695894499975,	0.0001312519434236,
>> -0.0000079156628396,	0.0001423655606105,	0.0000044733483182,
>> 0.0000303832556655,	0.0000577624190434,	0.0001193435284164,
>> 0.0002422477575812,
>> 0.0000216245975292,	0.0000091332643423,	-0.0000300982253054,
>> 0.0000992731601466,	-0.0000426583313588,	0.0001641146317929,
>> 0.0000311621614693,	-0.0000147821020927,	-0.0000767394607354,
>> 0.0000619936562782,	-0.0000306228761064,	0.0001495752154579,
>> 0.0000389317919640,	-0.0000714551280935,	-0.0000564616194935,
>> 0.0000384367900903,
>> 0.0000483253391811,	0.0000693354260457,	0.0000299178429478,
>> 0.0002685466918035,	0.0002067711081561,	0.0000311621614693,
>> 0.0003176493360736,	0.0002575792630182,	0.0001371966488704,
>> 0.0000436833885846,	0.0001442516276721,	0.0001075447728937,
>> 0.0000371155448252,	0.0000475873370276,	0.0002162409964174,
>> 0.0002870514043081,
>> 0.0000787580901352,	0.0000825217915015,	0.0000135672602503,
>> 0.0002580180069951,	0.0002695894499975,	-0.0000147821020927,
>> 0.0002575792630182,	0.0006217963583393,	0.0002368375072233,
>> 0.0000078625467985,	0.0002054774387807,	-0.0000066572248626,
>> 0.0000485854317294,	0.0002802199677114,	0.0001676465030622,
>> 0.0003028775764026,
>> 0.0000186474817658,	0.0000229122227269,	0.0001735431064391,
>> 0.0000725833959653,	0.0001312519434236,	-0.0000767394607354,
>> 0.0001371966488704,	0.0002368375072233,	0.0004475645060339,
>> -0.0000030389778729,	0.0001706183643212,	-0.0000017789896670,
>> 0.0000722657436668,	0.0001664088523103,	0.0001220193496918,
>> 0.0001641280878243,
>> 0.0000312260571354,	0.0000297662414650,	-0.0000133347388414,
>> 0.0000525639940758,	-0.0000079156628396,	0.0000619936562782,
>> 0.0000436833885846,	0.0000078625467985,	-0.0000030389778729,
>> 0.0000822356406019,	-0.0000226786278360,	0.0000752056105897,
>> 0.0000399801889185,	-0.0000441549693477,	0.0000047887593401,
>> 0.0000352165734549,
>> 0.0000217594093734,	0.0000334443258658,	0.0001387582890571,
>> 0.0001785049461665,	0.0001423655606105,	-0.0000306228761064,
>> 0.0001442516276721,	0.0002054774387807,	0.0001706183643212,
>> -0.0000226786278360,	0.0004304869804941,	0.0001566983136020,
>> 0.0000332770114864,	0.0000012432094922,	0.0002491186667930,
>> 0.0001285479414542,
>> 0.0000536298150897,	0.0000273254534933,	0.0000964898243724,
>> 0.0001781339191317,	0.0000044733483182,	0.0001495752154579,
>> 0.0001075447728937,	-0.0000066572248626,	-0.0000017789896670,
>> 0.0000752056105897,	0.0001566983136020,	0.0005292416268831,
>> 0.0000893358436932,	-0.0001009559617338,	0.0000888461032129,
>> 0.0000714719761291,
>> 0.0000166202592455,	0.0000202062301763,	-0.0000149571346672,
>> 0.0000597631329497,	0.0000303832556655,	0.0000389317919640,
>> 0.0000371155448252,	0.0000485854317294,	0.0000722657436668,
>> 0.0000399801889185,	0.0000332770114864,	0.0000893358436932,
>> 0.0001844874143317,	0.0000549019705905,	-0.0000117658984941,
>> 0.0000394986211508,
>> 0.0000451975061637,	0.0000026260702295,	0.0000104437939143,
>> 0.0000201160486244,	0.0000577624190434,	-0.0000714551280935,
>> 0.0000475873370276,	0.0002802199677114,	0.0001664088523103,
>> -0.0000441549693477,	0.0000012432094922,	-0.0001009559617338,
>> 0.0000549019705905,	0.0006686123611712,	-0.0001115788528761,
>> 0.0000151312169512,
>> -0.0000120364862228,	0.0000558975248740,	0.0001246900353191,
>> 0.0002582267884874,	0.0001193435284164,	-0.0000564616194935,
>> 0.0002162409964174,	0.0001676465030622,	0.0001220193496918,
>> 0.0000047887593401,	0.0002491186667930,	0.0000888461032129,
>> -0.0000117658984941,	-0.0001115788528761,	0.0014937840813054,
>> 0.0001299625832782,
>> 0.0000497117714376,	0.0000953647537111,	-0.0000171884354549,
>> 0.0002473268250781,	0.0002422477575812,	0.0000384367900903,
>> 0.0002870514043081,	0.0003028775764026,	0.0001641280878243,
>> 0.0000352165734549,	0.0001285479414542,	0.0000714719761291,
>> 0.0000394986211508,	0.0000151312169512,	0.0001299625832782,
>> 0.0004355778955394), c(16,16))
>>
>> covar  <- function(x) return (t(x) %*%mat %*% (x))
>> covargr <- function(x) return ( 2*mat %*% (x))
>>
>> upper1 = c(1, 0, 0, 0,  0, 0, 0,0 ,0 , 0, 0, 0, 0, 0, 0, 0)
>> lower1 = c(1, -0.25, -0.25, -0.25, -0.25, -0.25, -0.25, -0.25, -0.25,
>> -0.25,
>> -0.25, -0.25, -0.25, -0.25, -0.25, -0.25)
>> lower[1] = 1
>> init = upper1- 1/15.0
>> init[1] = 1
>>
>>  optim( init,  covar, covargr, method = "L-BFGS-B", lower = lower1, upper
>> =
>> upper1)
>> u1 <- c(0, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 0,
>> -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1)
>> u2<- array(u1, c(16, 2))
>> u2[,2] <- -u2[,2]
>>
>> u1<- t(u1)
>> c1<- array(c(-1, -1),c(2,1))
>>  constrOptim (init, covar, covargr, t(u2), c1, mu = 1e-054,  method =
>> "L-BFGS-B", lower = lower1, upper = upper1, outer.iterations = 100,
>> outer.eps = 1e-05)
>>
>>
>> The problem I face is an error message telling me that the method
>> "L-BFGS-B"
>> needs finite values of fn... 
>> while everything works well in the optim case without the inequalities.
>>
>> Does anybody have any clue about what might have gone wrong?
>>
>> THank you for your help
>>
>>
>> Georges
>>
>>
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 
> 

-- 
View this message in context: http://www.nabble.com/constrOptim-with-method-%3D-%22L-BFGS-B%22-tp17916460p17961273.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list