Feature: start values for optimization problem

[stelmo]

Sometimes it would be nice to help the solver by supplying the first guess manually

[exaexa]

what's the underlying JuMP API?

[exaexa]

(btw again this might be better directly in CTs)

[stelmo]

In JuMP you just do something like set_start_value(variable_ref, number). See here: https://jump.dev/JuMP.jl/stable/manual/variables/#Start-values. Yep, might be better in CTs to be more general

[exaexa]

edit here: not in CTs, CTs don't know anything about jump :D

[exaexa]

@stelmo btw is there a good usecase for this? So that I can write a test or so.

[stelmo]

A good use case would be to speed up FVA, because the solutions tend to be similar to each other

[exaexa]

FVA uses the single-model screen which already retains the previous solution. Maybe resetting the solution back would help but again that's not "exposed" to user. This would be only to supply e.g. a feasible solution before solving a (big) FBA or so.