Eric Hanson

Ah cool, I hadn't seen NEOS.jl!

Thanks for bringing this to my attention. This currently doesn’t work in Convex because Convex rewrites the problem to have a linear objective function subject to “second order cone constraints”....

> What is the current status on this ? same as before- Convex reformulates all problems with extended formulations (always) and passes a single variable objective. > Will it be...

That’s awesome! Getting some more tests for dual variables would be really great. Something I realized since making this issue is that Convex only supports dual variables for a few...

Step 1: remove `norm(A, p)` for matrices at all, so it produces an error. This was done in v0.14.0 via https://github.com/jump-dev/Convex.jl/pull/412. Step 2: re-introduce `norm(A, p)` as `norm(vec(A), p)`. This...

The functions `fix!` and `free!` can be used for this as well. You can `fix!` a variable to a value (`fix!(x, v)`), then solve the problem. Then you can `fix!`...

In this PR, I introduce a structure ```julia struct VectorAffineFunctionAsMatrix{M,B,V} matrix::M vector::B variables::V end ``` to represent an `MOI.VectorAffineFunction` in a way more conducive to composing affine functions of the...

I've pushed a commit that implements the ideas from the previous comment (and then one that fixes perf bugs), which improves the benchmarks quite a bit: ```julia julia> include("testproblem\\testproblem.jl") ┌...

To achieve type stability with this approach (or most others), we will need to make atoms parametric on the number of dimensions, because we often have different implementations for vectors...

The latest commit adds the unregistered dependency https://github.com/ericphanson/MatrixChainMultiply.jl to look at the impact of using that method for choosing how to evaluate the `VAFTape` objects. In the test problem it...