Constraint not satisfied yet result is optimal

1

Hello,
I am trying to use CVXR to solve my optimization problem, the code I have written is :

w = Variable(nrow(sigma))
n = Int(nrow(pweights))
objective = Minimize(quad_form(w,sigma-diag(specRisk^2)))
constraints1 = list(w+maxTradeSize*n >= pweights$pctEquity, #Constraint 1
                   w-maxTradeSize*n <= pweights$pctEquity,  #Constraint 2
                   w*sign(pweights$pctEquity) >= 0,         #Constraint 3
                   sum_entries(n)<=numTrades,               #Constraint 4
                   sum_entries(sign(pweights$pctEquity)*w)<=1.01*grossExposure, #Constraint 5
                   sum_entries(sign(pweights$pctEquity)*w)>=.99*grossExposure,  #Constraint 6
                   quad_form(w,sigma) <= volTarget,        #Constraint 7
                   n<=1, #Constraint 8
                   n>=0) #Constraint 9
problem = Problem(objective, constraints1)
result <- psolve(problem, MAXIT=as.integer(2000))
result$status

This gives me the status as optimal but when I check the output Constraint 7 is not satisfied at all.
I Tried the following code with lesser constraints (Here constraint 7 from the previous code has been made 1st) :

w = Variable(nrow(sigma))
n = Int(nrow(pweights))
objective = Minimize(quad_form(w,sigma-diag(specRisk^2)))
constraints2 = list(quad_form(w,sigma) <= volTarget, #Constraint 1
                   sum_entries(sign(pweights$pctEquity)*w)<=1.01*grossExposure, #Constraint 2
                   sum_entries(sign(pweights$pctEquity)*w)>=.99*grossExposure)  #Constraint 3
problem = Problem(objective, constraints1)
result <- psolve(problem, MAXIT=as.integer(2000))
result$status

This gives me an optimal output with the constraint being satisfied. I can not seem to figure out what the issue is.
Any help on this would be highly appreciated.

2

This is not a forum for CVXR in R but for CVX, a tool for MATLAB, so wrong place. You should try on https://github.com/cvxgrp/CVXR or directly in the support of the solver you are using underneath.

(I can already tell you that you will be asked for a fully reproducible code including all input data, log outputs etc. Otherwise it is impossible to answer. Moreover “constraint is not satisfied at all” is not so informative. By how much is it violated? Isn’t the violation just by a small \varepsilon that is perfectly normal within numerical tolerances? Could be numerical issues. Etc…)