Test quadratic code on Travis...

[IainNZ]

... by implementing a QCQP MPB interface to Ipopt, probably.

[joehuchette]

Tagging as 0.6, I think this should be a (fairly) high priority.

[mlubin]

Agreed. This includes quadratic MPS writer.

[mlubin]

Ipopt now supports QP (but not QCQP).

[mlubin]

I'm wondering how crazy it would be to translate quadratic SOC constraints x'x <= t^2 to the convex form x'x/t - t<= 0 before giving them to ipopt.

[joehuchette]

How well would Ipopt handle the case where t is very small (or zero)?

[mlubin]

Not well at all.

[joehuchette]

Well enough for testing, if we choose problems carefully?

[mlubin]

If we implement it, it'll probably have to be a solver option. Anyway we can probably get the tests to pass even if we give Ipopt the nonconvex form.

[mlubin]

We still need tests for the quadratic MPS writer, but untagging for 0.7.

[joehuchette]

Is this done yet? What's the status of the quadratic MPS writer?

[mlubin]

Still untested
On Jun 30, 2015 12:14 PM, "Joey Huchette" notifications@github.com wrote:

Is this done yet? What's the status of the quadratic MPS writer?

—
Reply to this email directly or view it on GitHub
#227 (comment).