[R] Problem with function gp in package cccp

[Jeff Reichman]

So its that gp() wouldn't even activate equality constraints unless it also
saw inequality ones. Interesting


-----Original Message-----
From: R-help <r-help-bounces using r-project.org> On Behalf Of
koller using science.iwi.ac.at
Sent: Friday, August 1, 2025 10:40 AM
To: r-help using r-project.org
Subject: Re: [R] Problem with function gp in package cccp

Thank you, Jeff and Bert!

1. As to the unconstrained problem I gave (which is not my main concern) I
should have added that in geometric programming it is assumed that the
target variables are positive, thus x,y > 0. This would also rule out my
given solution, (0, 0), of the unconstrained problem. The positivity
constraint in geometric programming is a consequence of the role the
logarithmic transformation plays in the theory of geometric programming.

2. However, my concern is to find out how to define equality constraints and
submit them as an argument to gp(), so I wanted to try out the most simple
case. But it seems as gp() implicitely assumes there must be at least
inequality constraints. If I add inequality constraints, then my equality
constraints are not ignored by gp(). So, in the following example, I added a
non-binding inequality constraint, and I receive the correct solution, this
time for the correct reasons:

## Minimise P + Q subject to P*Q = 1 and 0.5*P^(-1)*Q^(-1) <= 1
F0 <- diag(2)                       #  [m0 * n] m0 = number of terms of 
the posynomial, n = number of variables
g0 <- log(matrix(1,2,1))            #  [m0 * 1]
bexp <- 1
A <-  matrix(c(1,1), 1, 2)          #  one constraint, one row
b <-  log(matrix(bexp, 1, 1))
F1 <- matrix(c(-1,-1), 1, 2)        #  additional (redundant) inequality 
constraint [m1 * n] m1 = number of terms of the posynomial
g1 <- log(matrix(0.5*bexp, 1, 1))   #  [m1 * 1]
RES <- gp(F0, g0, FList=list(F1), gList=list(g1), A=A, b=b) sol <- RES$pdv$x
c(sol)  # [1] 1 1
RES <- gp(F0, g0, FList=list(F1), gList=list(g1))  # without the equality
constraint sol <- RES$pdv$x
c(sol)  # [1] 0.7071069 0.7071069

Best regards,

Wolfgang






Am 2025-08-01 04:18, schrieb Jeff Reichman:
> In gp(F0, g0, A = A, b = b) : No restrictions provided, trying solve()
> 
> 
> This likely means that while you've specified A and b, the solver 
> doesn't recognize any inequality (<=) or equality (=) restrictions - 
> because in the context of the gp() function from the CVXR or similar 
> geometric programming libraries in R, you're expected to provide 
> explicit inequality/equality markers or properly formatted constraint 
> matrices.
> 
> So what's missing? Probably an argument like ineq = ... or eq = ... to 
> declare the type of constraint.
> 
> Here's how to express your constraint properly, assuming the library 
> expects inequalities or equalities to be marked explicitly:
> 
> # Properly declare equality constraints gp(F0, g0, eq=A, eq.b=b)
> 
> Or possibly (depending on the package):
> 
> gp(F0, g0, Aeq=A, beq=b)
> 
> If the solver expects inequality constraints:
> 
> gp(F0, g0, A=A, b=b, meq=1)  # Indicating 1 equality constraint
> 
> Jeff Reichman
> 
> 
> -----Original Message-----
> From: R-help <r-help-bounces using r-project.org> On Behalf Of 
> koller using science.iwi.ac.at
> Sent: Wednesday, July 30, 2025 2:32 PM
> To: r-help using r-project.org
> Subject: [R] Problem with function gp in package cccp
> 
> Dear all!
> 
> I tried to solve a simple test example of a geometric program:
> minimize x + y   s.t.   xy = 1
> The solution should be (1, 1).
> 
> I tried:
> F0 <- diag(2)                  #  [m * n] m = number of terms of the
> posynomial, n = number of variables
> g0 <- log(matrix(1,2,1))       #  [m * 1]
> A <-  matrix(c(1,1), 1, 2)     #  one constraint, one row
> b <-  log(matrix(1,1,1))
> RES <- gp(F0, g0, A=A, b=b)
> 
> Though I received the correct solution, the function gives a warning:
> In gp(F0, g0, A = A, b = b) : No restrictions provided, trying solve().
> 
> Also, when I try
> RES <- gp(F0, g0, A="A", b=b)
> or
> gp(F0, g0)
> 
> receiving the same result, I find out, that my equality constraints 
> are not considered at all.
> 
> In my view, the unconstraint problem should have (0, 0) as a solution, 
> btw.
> 
> Am I missing something fundamental?
> 
> Thanks in advance!
> 
> Wolfgang
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see 
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> https://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

______________________________________________
R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
https://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.