[R] multi dimensional optim problem
Berend Hasselman
bhh at xs4all.nl
Tue Dec 18 22:15:33 CET 2012
Forgot to cc the help list.
On 18-12-2012, at 19:40, Lu, James T wrote:
> I am attempting to use optim to solve a neural network problem. I would like to optimize coefficients that are currently stored in a matrix
>
> Y=270 x 1
> X= 27- x 14
> b1= 10x14
> b2= 11x1
> V= 10 x 14 set of prior variances.
>
> I have the following function:
>
> posterior.mode1=function(y,X,b_0,b2,V) {
>
> log.like=function(b1) {
> a_g=compute(b1)
> z_g=tanh(a_g);
> z_g=cbind(1,z_g)
> p=softmax(z_g%*%b2);
> a=sum(y*log(p)+(1-y)*log(1-p));
> return(a);
> }
>
> compute=function(b1) {
> a_g=NULL;
> for(i in 1:nrow(b1)){
> a_g=cbind(a_g,X%*%b1[i,])
> }
> return(a_g);
> }
>
> log.posterior=function(b1) {
> -log.like(b1)+1/2*t(as.vector(b1))%*%diag(as.vector(V))%*%as.vector(b1)
> }
>
> a=optim(b_0,log.posterior,method="CG",hessian=TRUE)
> return(a);
> }
>
> When I run
>
> posterior.mode1(y,X,b1,b2,b1)
>
>
>
> I get the following error
>
>
>
> Error in 1:nrow(b1) : argument of length 0
optim treats parameter b_0 as a vector and passes as such to the function it is trying to optimize.
So in log.like you should convert b1 back to a matrix of the correct dimensions like so
b1 <- matrix(b1,nrow=10)
Then it should work.
Note: you haven't provided function softmax.
Berend
More information about the R-help
mailing list