fixed Dimension for mdata.Phi array

[Rym28]

Hi,

I would like to run an ILP model with three objectives.

the problem in the types.jl, the Phi is fixed at two dimensions, as using WeightedSum method, i can deal with more than two objectives,

So, I could not check the value of the third dimension in Phi.

thanks

[anriseth]

Phi is a matrix of size #objectives times #objectives. Column i represents the values of all the objectives at the minimizer of objective i.

So the values of the third objective area in row three.

[Rym28]

Yes , I get the idea?

Unfortunatly, that for the nadir and utopia points are not developped in the MultiJuMP package, which represent important informations for the generated Pareto front set.

[anriseth]

You can extract the Nadir and Utopia points from Phi: https://github.com/anriseth/MultiJuMP.jl/blob/07474f7e1bd8e4066f8de92430c1f4b869745f56/src/linear.jl#L30-L31 Just remember to adjust for whether you are minimizing or maximizing an objective to get the correct signs of Fmin and Fmax.

I don't understand what this has to do with making the dimension of Phi equal to the number of objectives though?

[Rym28]

I fthough that Phi is always fixed to 2D as mentionned in types.jl mutable struct MultiData{Tx,To} objectives::Vector{SingleObjective} f1::SingleObjective f2::SingleObjective normalf1 normalf2

linear::Bool
# Number of points in each direction
# of Pareto submanifold
pointsperdim::Int
#
# stored values
utopiavarvalues::Vector{Vector{Tx}}
utopia::Vector{To}
nadir::Vector{Tx}
paretovarvalues::Vector{Vector{To}}
paretofront

Phi::Array{To,2}
Fstar::Vector{To}
inequality::Bool

end

Also, writing mdata.utopia, i got empty array.

Otherwise, I'm ooking to get the knee point: by finding the solution which has the best trade-off between objectives amon all the generated non-dominated solutions.

thanks

thanks

[anriseth]

I fthough that Phi is always fixed to 2D as mentionned in types.jl

Yes, Phi is a matrix, that is, a 2-dimensional Array, but the size (number of rows and columns) is equal to the number of objectives.

Also, writing mdata.utopia, i got empty array.

Yes, that was an old field that was not in use anymore. I've added some functionality back in #34.

Otherwise, I'm ooking to get the knee point: by finding the solution which has the best trade-off between objectives amon all the generated non-dominated solutions.

Maybe I have misunderstood you, but the "best" trade-off sounds rather subjective when you have multiple objectives?

[Rym28]

Thanks for these clarifications.

Otherwise, after generating multiple non-dominated solutions, we have to help the Decision maker which is solution is more closer to his preferences. This vector or solution is the Knee point in the litreature.

[anriseth]

we have to help the Decision maker which is solution is more closer to his preferences.

Feel free to submit a PR to support such functionality ;)