JuliaOpt/JuMP-dev

Developer-focused chat about JuMP, M̶a̶t̶h̶P̶r̶o̶g̶B̶a̶s̶e̶, MathOptInterface, solver interfaces, and JuMP extensions. Towards JuMP 1.0! Basic usage questions will be answered more quickly at https://discourse.julialang.org/c/domain/opt.

Oscar Dowson

@odow

Good to know re documentation. Let me know if you have suggestions for how to improve it further

Oscar Dowson

@odow

@/all Two items:

We're postponing tomorrow's monthly call a week until June 3. Same time, 2 pm Eastern. Please ping me if you would like to attend and need a calendar invite.
Please remember to get your JuMP-dev abstracts in by May 31: https://jump.dev/meetings/juliacon2021/. Given the tight timeframe, there will not be an extension!

Miles Lubin

@mlubin

Not only should you submit abstracts, you should also tell your friends, colleagues, and twitter followers to submit abstracts :)

Miles Lubin

@mlubin

@/all All JuMP contributors are invited to participate in this survey conducted by NumFOCUS: https://discourse.julialang.org/t/jump-numfocus-contributor-survey/62141

Oscar Dowson

@odow

@/all a reminder that May's (delayed) monthly developer call is tomorrow, Thursday June 3 at 2pm Eastern. Ping me or @mlubin for an invite if you don't already have one.

freemin7

@freemin7

Is there any known way in the literature to deal with optimization variables grows when some decisions are taken: For example planning and optimizing the collonisation of the universe with self replicating probes (a probe can be instructed to build housing for collonists/begin terraforming or strip the planets resources to some degree and create a variable number of probes that take new planets.
Precreating all variables would be ludicrous.

freemin7

@freemin7

Other examples would be a robot using optimization to plan to cut a veggie (each cut increases the state space) or approximating static structures out of divisible material. Or optimizing a 4-X style game (build dwellings to hire to units using resources to ... ) play. Or optimizing placement to cause/prevent a kessler syndrome in orbit.
If someone has a name for such problems that would be helpful too.

Miles Lubin

@mlubin

You can look at cutting stock problems as a starting point

Oscar Dowson

@odow

https://jump.dev/JuMP.jl/stable/tutorials/Mixed-integer%20linear%20programs/cutting_stock_column_generation/

freemin7

@freemin7

So there is no formal language to express such problems and one just writes some custom code around a (MI)(N)LP solver to somehow get a handle on it?

Miles Lubin

@mlubin

The formal tools are column generation/configuration LPs/dynamic programs

At least in the MIP world.

freemin7

@freemin7

so no formal modeling language that includes the notion of generating new variables, ok.

Miles Lubin

@mlubin

Not that I'm aware of

You might also want to look at warren powell's sequential decision analytics

"optimization variables grows when some decisions are taken" implies some kind of sequential decision making

freemin7

@freemin7

This sounds quiet like what i am after. Does Julia have any solvers for that?

Miles Lubin

@mlubin

There are no solvers :)

There's a book (which I haven't read) that provides a mathematical framework

freemin7

@freemin7

Mhh, the toy problem is in continuous time, but the framework looks useful.

BridgingBot

@GitterIRCbot

[slack] <oxinabox> I have seen genetic algorithms used for that. But that had no optimality garentees.

[slack] <oxinabox> It feels like it maybe can be done with some careful dynamic programming?

[slack] <oxinabox> But neither of those are preboxed solutions

Jean-Francois Baffier

@Azzaare

Hi there! I am writing an article for LocalSearchSolvers.jl (and CBLS.jl its MOI/JuMP interface) and I want to cite both MOI and JuMP. In the MOI repo, the ref point to an (in press) article. Is it still in press or is there any update on this part ? I can site the arxiv paper too anyway, but I'm just trying to be accurate!
Also, @Wikunia @dourouc05 , I plan to site both of your packages. Any related paper that I might be missing?

For ConstraintSolver and CPLEXCP I refer to the repo at the moment!

Oscar Dowson

@odow

It's still in press... :(

Jean-Francois Baffier

@Azzaare

Well, I will still refer it and the arxiv version. When I have to submit the final version I will check with you again ;)

Ole Kröger

@Wikunia

No paper about it but looking forward to the article 😊

Jean-Francois Baffier

@Azzaare

It is a modest WIP article for a workshop. But journal version will come eventually ^^'. Thanks for the answer

BridgingBot

@GitterIRCbot

[slack] <matbesancon> in discrete time, you might want to look at multi-stage stochastic optimization, this is essentially what you are describing if I understand correctly

Thibaut Cuvelier

@dourouc05

Also, @Wikunia @dourouc05 , I plan to site both of your packages. Any related paper that I might be missing?

Hi!
I have no paper you can cite for now (not even on arXiv). However, I have just created a Zenodo DOI so you have something to link to: https://zenodo.org/record/4908782 (it is now in the README for this package). I don't plan to have anything specific for CPLEXCP, though (similarly, you have publications for JuMP, MPB, and MOI, but not for the individual solvers).
Looking forward to your publication :)!

aaaaluba

@aaaaluba

@odow I have a mistake on gurobi : could not load library "gurobi91"The specified module could not be found. It happened rencently. Never saw it before.

aaaaluba

@aaaaluba

Oh, I fixed that via jump-dev/Gurobi.jl#387 @odow

Thanks

BridgingBot

@GitterIRCbot

[slack] <chris.peel> A year or so ago I don't think there were any integer programming solvers that could handle BigInts or BigFloats. Is that still the case?

BridgingBot

@GitterIRCbot

[slack] <matbesancon> In general or in MOI? For the latter, I'd still say no. For the former, SCIP has an exact mode (not BigFloat but rationals)

Joshua Pulsipher

@pulsipher

I ran into linkchecks failing when I build docs that involved Ipopt.jl on GitHub Actions with ubuntu-latest. It turns out the importing Ipopt.jl breaks curl because of LD_LIBRARY_PATH that Ipopt uses. See the discussion: https://discourse.julialang.org/t/link-checking-with-github-actions-not-working/61823. I made a minimum working example that explains the problem: https://github.com/pulsipher/JuliaCurlTest.

BridgingBot

@GitterIRCbot

[slack] <chris.peel> Both of you answers are useful. I'll look at SCIP. Thanks!

Oscar Dowson

@odow

There is an open issue in Ipopt to fix that

BridgingBot

@GitterIRCbot

[slack] <chris.peel> I looked at SCIP.jl; I was able to get some simple, non-BigInt examples working, but no the BigInt examples. I guess it's not as simple as I'd hoped

Oscar Dowson

@odow

Why do you need big ints?

Juan Pablo Vielma

@juan-pablo-vielma

does Gurobi.jl currently support duals for soc constraints?

Oscar Dowson

@odow

Not sure. Don't think so

Robert Schwarz

@rschwarz

I don't think that the exact-SCIP functionality is available through SCIP.jl.

Mathieu Besançon

@matbesancon

No it's not indeed. It won't be worth integrating before the next major release

BridgingBot

@GitterIRCbot

[slack] <matbesancon> the bridge seems dead again but as noted in the gitter chat, the exact SCIP part that uses extended integers and rationals is not exposed in the SCIP.jl wrapper

Chris

@christianpeel

@odow I have a toy subset sum solver that can work with BigInts, and wanted to compare it with a generic solver in the JuMP ecosystem using BigInts. It's not critical.

BridgingBot

@GitterIRCbot

[slack] <ericphanson> JuMP only supports Float64 anyway, though one can use MOI directly (or Convex.jl for some problems)

Lea Kapelevich

@lkapelevich

i found this unexpected after initializing a Vector{AffExpr} with zeros:

julia> using JuMP; m = Model(); @variable(m, x);

julia> A = zeros(AffExpr, 2); add_to_expression!(A[2], 1, x); A
2-element Vector{AffExpr}:
 x
 x

i was expecting the same as:

julia> B = AffExpr[0, 0]; add_to_expression!(B[2], 1, x); B
2-element Vector{AffExpr}:
 0
 x

BridgingBot

@GitterIRCbot

[slack] <mlubin> jump-dev/JuMP.jl#1151

Where communities thrive

People

Repo info

Activity