Augmented EpsilonConstraint

[odow]

Our EpsilonConstraint method is

min f_1(x)
      f_2(x) <= eps

we might consider

min f_1(x) + p * f_2(x)
      f_2(x) <= eps

for some small p.

[kofgokhan]

There can be an option for either using augmentation or two-stage formulation.

[odow]

Yeah since we currently require 2 objectives, we could add a new parameter for controlling p that defaults to 0.0?

[xgandibleux]

Good suggestion.

[odow]

Is this really needed? Our current method finds the extra points, then filters them out. I can see that it might be more efficient if you had lots of non-dominated points that were not efficient. But there's also a cost to doubling the number of solves for each point, or making the numerics worse by adding a parameter.