Berend Hasselman <bhh <at> xs4all.nl> writes: > It seems you are absolutely right. I always assumed a quadratic programming solver will -- as all linear programming solvers do -- automatically require the variables to be positive.
I checked it for some more examples in 10 and even 100 dimensions, and the results now agree. Still, it's a bit disappointing that 'quadprog' will not solve problems with 10 points in R^3, because the corresponding matrices are not positive definite. Thanks Hans Werner > > After having a closer look at this problem, I believe you did not include > the constraint x_i >= 0 in the call to solve.QP. > So with this modification of your code > > A <- matrix(rep(1,3),nrow=4,ncol=3,byrow=TRUE) > A[2:4,] <- diag(3) > b <- c(1,0,0,0) > > sol3 <- solve.QP(D, d, t(A), b, meq = 1) # first row of A is an equality > sol3 > p0 <- c(C %*% sol3$solution) > r0 <- sqrt(-sol3$value) > p0 > r0 > sqrt(colSums((C - p0)^2)) > > one gets the correct answer. > BTW LowRankQP seems to postulate x_i >=0 if I read its manual correctly. > > Berend ______________________________________________ R-help@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.
