getobjbound for conic_to_LPQP

[chriscoey]

@mlubin, in new QP tests

QP maximize: Error During Test
  Got an exception of type MethodError outside of a @test
  MethodError: no method matching getobjbound(::MathProgBase.SolverInterface.ConicToLPQPBridge)
  Closest candidates are:
    getobjbound(!Matched::GLPKMathProgInterface.GLPKInterfaceMIP.GLPKMathProgModelMIP) at /home/coey/.julia/v0.5/GLPKMathProgInterface/src/GLPKInterfaceMIP.jl:549
    getobjbound(!Matched::Cbc.CbcMathProgSolverInterface.CbcMathProgModel) at /home/coey/.julia/v0.5/Cbc/src/CbcSolverInterface.jl:190
    getobjbound(!Matched::Mosek.MosekMathProgSolverInterface.MosekLinearQuadraticModel) at /home/coey/.julia/v0.5/Mosek/src/MosekLPQCQPInterface.jl:661

[chriscoey]

also one test isn't passing there because the answer is a factor of 2 off. it could be caused by maybe forgetting the 2 in the RSOC?

getobjectivevalue(m) = 8.022765808609499
QP large (cardls): Test Failed
  Expression: isapprox(getobjectivevalue(m),16.045564,atol=TOL)

[mlubin]

Try out JuliaOpt/MathProgBase.jl#160 for the getobjbound issue.

With

@variable(m, w == 2)
@constraint(m, sum(t.^2) <= u*w)

You get u >= sum(t.^2)/2. Is that what you want?

[chriscoey]

Yes, I wrote that part correctly.

…

On Mar 29, 2017 10:46 AM, "Miles Lubin" ***@***.***> wrote: Try out JuliaOpt/MathProgBase.jl#160 <JuliaOpt/MathProgBase.jl#160> for the getobjbound issue. With @variable(m, w == ***@***.***(m, sum(t.^2) <= u*w) You get u >= sum(t.^2)/2. Is that what you want? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJq0k6hJnJOFk6BSqVnGhCWH_MI-3Yfdks5rqm63gaJpZM4Msim6> .

[mlubin]

So the objective is to minimize sum(t.^2)/2 + rho*sum(x.^2)/2?

[chriscoey]

Yes

…

On Mar 29, 2017 10:53 AM, "Miles Lubin" ***@***.***> wrote: So the objective is to minimize sum(t.^2)/2 + rho*sum(x.^2)/2? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJq0kxHwia_ndsawslpDKzPm4FCtZr_Tks5rqnBzgaJpZM4Msim6> .

[mlubin]

The constraints are being enforced as written:

julia> sum(getvalue(t).^2)/2 + rho*sum(getvalue(x).^2)/2
8.022766055303913

julia> getobjectivevalue(m)
8.022766159395793

[chriscoey]

Then that is the correct objective. But the conic tests pass with objective twice that, suggesting an issue with conic to lpqp On Mar 29, 2017 11:00, "Miles Lubin" <notifications@github.com> wrote: The constraints are being enforced as written: julia> sum(getvalue(t).^2)/2 + rho*sum(getvalue(x).^2)/2 8.022766055303913 julia> getobjectivevalue(m) 8.022766159395793 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJq0k7GUtmAbOyRYbzRCLvH4wJlnF9X-ks5rqnH7gaJpZM4Msim6> .

[mlubin]

That suggests an issue with the conic formulation, not conic to lpqp.

[chriscoey]

I don't understand. There is only one formulation.

…

On Mar 29, 2017 11:12 AM, "Miles Lubin" ***@***.***> wrote: That suggests an issue with the conic formulation, not conic to lpqp. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJq0kxmptIT-8tUmquJWmHfAP_Qf1HV8ks5rqnTqgaJpZM4Msim6> .

[mlubin]

What solver is reporting 16.045564?

[chriscoey]

All. Could be a bug in pajarito too.

…

On Mar 29, 2017 11:17 AM, "Miles Lubin" ***@***.***> wrote: What solver is reporting 16.045564? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJq0k6zCUnvPmvqbUckAcsZE1fPHKy8Iks5rqnYAgaJpZM4Msim6> .

[mlubin]

I was testing with mosek above through conic_to_lpqp and it was giving the right answer.

…

On Mar 29, 2017 11:25, "Chris C." ***@***.***> wrote: All. Could be a bug in pajarito too. On Mar 29, 2017 11:17 AM, "Miles Lubin" ***@***.***> wrote: > What solver is reporting 16.045564? > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#338 issuecomment-290123188>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/ AJq0k6zCUnvPmvqbUckAcsZE1fPHKy8Iks5rqnYAgaJpZM4Msim6> > . > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#338 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABp0M7jBBbqaz225upLi1_OcWvZ9M7j3ks5rqnfNgaJpZM4Msim6> .

[chriscoey]

OK this seems to be fixed.

Tests passing when on MPB master.