Problems when moving plb & pub

[ajlahwh]

Hello,

I guess there might be some potential problems when moving plb & pub.

The initialization of one parameter in my model: lb: 6.14421235e-06 ub: 1.62754791e+05 plb: 0.04978707 pub: 20.08553692 X0: 10.067662

It then says: bads:InitialPointsTooClosePB: The starting points X0 are on or numerically too close to the hard bounds lb and ub. Moving the initial points more inside...

Now the value becomes: X0: 162.75479756

It then says: bads:TooCloseBounds: For each variable, hard and plausible bounds should not be too close. Moving plausible bounds.

Now the value becomes: LB_eff: 1.62754798e+02 UB_eff: 1.62592037e+05 plb: 1.62754798e+02 pub: 20.08553692

It then says: bads:InitialPointsOutsidePB. The starting points X0 are not inside the provided plausible bounds plb and pub. Expanding the plausible bounds... ValueError: bads:StrictBounds: For each variable, hard and plausible bounds should respect the ordering lb <= plb < pub <= ub.

Your help is greatly appreciated. Thanks!

[lacerbi]

Thanks, I will look into this at some point. However, as a basic recommendation, it seems that your variables would be best specified in log scale, as they span 11 orders of magnitude (!), from 10^-6 to 10^5. This is almost surely what's causing the issue here.

[ajlahwh]

Thanks, I will look into this at some point. However, as a basic recommendation, it seems that your variables would be best specified in log scale, as they span 11 orders of magnitude (!), from 10^-6 to 10^5. This is almost surely what's causing the issue here.

Okay, I see. Thank you for the recommendation!