search

jump-dev / JuMP.jl

Modeling language for Mathematical Optimization (linear, mixed-integer, conic, semidefinite, nonlinear)
http://jump.dev/JuMP.jl/
Other
2.23k stars 394 forks source link

Tidy up quadratic constraint printing #35

Closed mlubin closed 10 years ago

joehuchette commented 11 years ago

What did you mean by this? All I can think of is grouping like variables (e.g. 2 xz + 3zx => 5xz).

IainNZ commented 11 years ago

Yeah that was it

joehuchette commented 10 years ago

Also:

julia> addConstraint(m,x^2-x^2==0)
1.0 x² - 1.0 x² == 0
IainNZ commented 10 years ago

Hah, yep. I'm still in love with those ²s :D

IainNZ commented 10 years ago

Gonna claim 0.3 as the quadratic constraints update - so I'll put this under 0.3

joehuchette commented 10 years ago

Just quadratic? Want to shoot for full nonlinear (Ipopt, Mosek) support as well?

IainNZ commented 10 years ago

So we'll definitely support quadratic for those, but I think general convex optimization depends on getting automatic differentiation done well first, so, later.

joehuchette commented 10 years ago

Fair enough. More structured stuff like SDPs would be really cool, though!

mlubin commented 10 years ago

CVX-style convex doesn't need AD afaik, the conic structure is built up with the expression. Replicating CVX will be a bit past 0.3 though :). General nonlinear does need AD, which might come sooner rather than later. I had a good chat with the AD guys at Argonne.