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

[lhaba]

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

-- 
View this message in context: http://www.nabble.com/constrOptim-with-method-%3D-%22L-BFGS-B%22-tp17916460p17916460.html
Sent from the R help mailing list archive at Nabble.com.