Optimisation
1
0
Entering edit mode
SimonNoël ▴ 450
@simonnoel-3455
Last seen 10.2 years ago
Hi every one, I have a big formula with some parameter that I want to minimise with some condition like sum of all of them must be one, in some condition, one of them must equal 0, etc. I am trying to solve that with constrOptim but I got some broblem... Like how to have a parametter equal to 0 Is there any package who is more user friendly that you can suggest to me? Simon No??l CdeC
• 1.0k views
ADD COMMENT
0
Entering edit mode
SimonNoël ▴ 450
@simonnoel-3455
Last seen 10.2 years ago
Hello, Anyone can help? Simon No??l CdeC _________________________________________________________________ De : Simon No??l Date d'envoi : 9 f??vrier 2012 13:37 ?? : bioconductor at r-project.org Objet : Optimisation Hi every one, I have a big formula with some parameter that I want to minimise with some condition like sum of all of them must be one, in some condition, one of them must equal 0, etc. I am trying to solve that with constrOptim but I got some broblem... Like how to have a parametter equal to 0 Is there any package who is more user friendly that you can suggest to me? Simon No??l CdeC
ADD COMMENT
0
Entering edit mode
you probably want to post this on r-help or perhaps look at the optimx package. you might also want to include a piece of code showing what precisely it is that you are trying to do. this is more likely to get a response than the fairly vague message below. hope it helps, --t 2012/2/22 Simon Noël <simon.noel.2@ulaval.ca> > > Hello, > > Anyone can help? > > Simon Noël > CdeC > _________________________________________________________________ > > De : Simon Noël > Date d'envoi : 9 février 2012 13:37 > À : bioconductor@r-project.org > Objet : Optimisation > Hi every one, > > I have a big formula with some parameter that I want to minimise with > some > condition like sum of all of them must be one, in some condition, one > of them must equal 0, etc. > > I am trying to solve that with constrOptim but I got some broblem... > Like > how to have a parametter equal to 0 Is there any package who is more > user friendly that you can suggest to me? > > Simon Noël > CdeC > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > -- *A model is a lie that helps you see the truth.* * * Howard Skipper<http: cancerres.aacrjournals.org="" content="" 31="" 9="" 1173.full.pdf=""> [[alternative HTML version deleted]]
ADD REPLY
0
Entering edit mode
I tought that big formula will scare some people... But I will try... So I have that function : variance <- function(w, result) { w22 = w[1] w21 = w[2] w12 = w[3] w11 = 1 - w12 - w21 - w22 ( ((1/2 * w22)^2 * result[1]) + ((w21)^2 * result[2]) + ((w12)^2 * result[3]) + ((2 * w11)^2 * result[4]) + (w22 * w21 * result[5]) + (w22 * w12 * result[6]) + (2 * w22 * w11 * result[7]) + (2 * w21 * w12 * result[8]) + (4 * w21 * w11 * result[9]) + (4 * w12 * w11 * result[10]) ) } I want to find the combinaison of value for my W that will give the best minimisation of my equation. Each w must be between 1 and 0. Sum of all om them must = 1. Until now, everything seem to be ok. I use constrOptim like that : Amat <- matrix(c(1,-1,0,0,0,0,0,0,1,-1,0,0,0,0,0,0,1,-1),6,3) Bvec <- c(0,-1,0,-1,0,-1) varLamda = constrOptim(theta=w, f=variance, grad=NULL, ui=Amat, ci=Bvec, result = result) result is defined as variance1 <- function(i22, i21, i12, i11) { result = vector(length = 10) result[1] = VLogRc(i22) result[2] = VLogRc(i21) result[3] = VLogRc(i12) result[4] = VLogRc(i11) result[5] = covlogi(i22, i21) result[6] = covlogi(i22, i12) result[7] = covlogi(i22, i11) result[8] = covlogi(i21, i12) result[9] = covlogi(i21, i11) result[10] = covlogi(i12, i11) result } and covlogi <- function(i1, i2) { i = vector(length = length(i1)) for(j in 1:length(i1)) { if(names(i1)[j] == names(i2)[j]) { i[j] = 1/i1[j] }else { i[j] = 0 } } sum(i) } VLogRc <- function(i) { ( (1/i[1]) + (1/i[2]) + (1/i[3]) + (1/i[4]) ) } my i are #i11 = (q11 * q00) / (q10 * q01) #i12 = (q12 * q00) / (q10 * q02) #i21 = (q21 * q00) / (q20 * q01) #i22 = (q22 * q00) / (q20 * q02) i11[1] = Q[5] i11[2] = Q[1] i11[3] = Q[4] i11[4] = Q[2] i12[1] = Q[6] i12[2] = Q[1] i12[3] = Q[4] i12[4] = Q[3] i21[1] = Q[8] i21[2] = Q[1] i21[3] = Q[7] i21[4] = Q[2] i22[1] = Q[9] i22[2] = Q[1] i22[3] = Q[7] i22[4] = Q[3] and each of my Q are the number of subkect with a particular combinaison of SNPs. It may happen that one of the Q = 0. Like if q22 = 0, we will have with VLogRc = ( (1/i[1]) + (1/i[2]) + (1/i[3]) + (1/i[4]) ) where i[1] = 0 so the result is INF and that's really bad. One way to avoid that is to write a new formula without w22.. And an other when a problem come for w21 and an other for w11 and one other for w12 and an other for w22 and w21, etc. There also a lot of if/else for each possible q combinaison with 0 value. So it's not a goot solution... Something simple is if I have an INF, and w = 0, ther the problem will be solved... If we take the exemple of q22 =0, w22 must = 0 in orther that the program work... But how to corce that with constrOptim? Is there a better fonction to solve my problem? Simon No??l CdeC _________________________________________________________________ De : Tim Triche, Jr. [tim.triche at gmail.com] Date d'envoi : 22 f??vrier 2012 16:11 ?? : Simon No??l Cc : bioconductor at r-project.org Objet : Re: [BioC] RE : Optimisation you probably want to post this on r-help or perhaps look at the optimx package. you might also want to include a piece of code showing what precisely it is that you are trying to do. this is more likely to get a response than the fairly vague message below. hope it helps, --t 2012/2/22 Simon No??l <[1]simon.noel.2 at ulaval.ca> Hello, Anyone can help? Simon No????l CdeC _________________________________________________________________ De : Simon No????l Date d'envoi : 9 f????vrier 2012 13:37 ????? : [2]bioconductor at r-project.org Objet : Optimisation Hi every one, I have a big formula with some parameter that I want to minimise with some condition like sum of all of them must be one, in some condition, one of them must equal 0, etc. I am trying to solve that with constrOptim but I got some broblem... Like how to have a parametter equal to 0 Is there any package who is more user friendly that you can suggest to me? Simon No????l CdeC _______________________________________________ Bioconductor mailing list [3]Bioconductor at r-project.org [4]https://stat.ethz.ch/mailman/listinfo/bioconductor Search the archives: [5]http://news.gmane.org/gmane.science.biology.informatics.conductor -- A model is a lie that helps you see the truth. [6]Howard Skipper References 1. mailto:simon.noel.2 at ulaval.ca 2. mailto:bioconductor at r-project.org 3. mailto:Bioconductor at r-project.org 4. https://stat.ethz.ch/mailman/listinfo/bioconductor 5. http://news.gmane.org/gmane.science.biology.informatics.conductor 6. http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf
ADD REPLY

Login before adding your answer.

Traffic: 652 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6