COBYLA reports solution out of bounds

[I-Fons]

Hi!

I am specifying the optimizaton options and calling the optimizer like:

# # OPTIMIZATION SPECIFICATIONS
# Guess values of the manipulated variables to initialize optimization
Q0 = 3. # mL/min or cm^3/min
t_load0 = 1. # min
optVar0 = [Q0 , t_load0]
# Algorithm
opt = StochOptim.Opt(:LN_COBYLA, 2)
grad = [] # Because COBYLA is gradient free
# Bounds 
opt.lower_bounds = [1e-6, 2.] # Lower bounds cannot be 0 because objective function returns NaN at some point
opt.upper_bounds = [5., 100000.]
# Stopping criteria
opt.xtol_rel = 1e-3
(minf_v1,minx_v1,ret_v1) = StochOptim.optimize_stoch_prob_v1(opt, grad , pCol , pLPM , c_in_vec , penalty_factor , optVar0, max_loses , risk)
StochOptim.print_optim_results(minx_v1,ret_v1,opt,c_in_vec, pCol , pLPM)`

But I get a solution that violates the lower bound:
(minf_v1,minx_v1,ret_v1) =(-38.6511199999541, [5.0, 0.5], :XTOL_REACHED)

Am I doing something wrong?

[mzaffalon]

StochOptim is not part of NLopt and some variables definitions (e.g. penalty_factor) are missing: could you please post a minimal workable example?

[odow]

Closing as stale and non-reproducible. Potentially also a duplicate of #166.

Please re-open if you have a reproducible example.