hfawzi / cvxquad Goto Github PK

Hi, sorry for the noise but I'm interested in knowing: (1) the dual cones of your recognized cones, and (2) the conjugate functions of your recognized convex functions. I didn't see this in the reference in the readme, so I'm wondering if you can point to any sources that might help me with this.

FYI @lkapelevich and I work on a generic conic interior point solver https://github.com/chriscoey/Hypatia.jl and we want to support some of these cones (and cones constructed from your functions). We think we can construct SC barriers for some cones (eg a matrix entropy perspective cone), and for others we can construct non-SC but probably "good-enough" barriers (eg matrix relative entropy cone).

Thanks!

Minimum reproducible example of trace_mpower with exponent t = 1 producing concave expression.

cvx_begin
variable A(2,2) semidefinite
trace_mpower(A,1)
ans =
    cvx concave expression (scalar)

In comparison:

trace(A)
ans =
    cvx real affine expression (scalar)

Of course, this bug can be circumvented by using trace instead of trace_mpower if t is known to be 1.

I try to run examples/rel_entr_entanglement.m, but get the following error.

>> rel_entr_entanglement
Undefined function or variable 'randRho'

I've searched the whole repo and cannot find its definition (similar issue for Tx). Could you please provide these functions? thx

I put CVXQUAD;s exponential.m in the CVX sets directory, CVXQUAD's Pade approximation was invoked for entr and rel_entr., as expected.

However, CVX's successive approximation method was used for log_det . I have not looked to see exactly which "exponential cone" inducing functions or circumstances use CVX's successive approximation method rather than CVXQUAD's Pade approximation.

I have been attempting to get this ported to Julia's Convex.jl. I believe the maintainer of Convex.jl (or maybe myself) has contacted you about this a while back. He was requesting that we cite the cvxquad license in the Convex.jl license file. Can you please add a license to your code? I believe that either BSD or MIT would be compatible with Convex.jl, though of course it's your software and your choice.

There is a bug in quantum_rel_entr which is triggered when 2nd argument is numeric but 1st argument is not.

Line 69
elseif cvx_isconstant(A) || cvx_isconstant(B)
of quantum_rel_entr.m should not have
|| cvx_isconstant(B)
because
cvx_optval = -quantum_entr(A,m,k) - trace_logm(B,A,m,k);
is triggered by satisfying this, is not correct for that case, and and will trip an error in trace_logm.

I leave it to you whether to add an
elseif cvx_isconstant(B)
with a following appropriate line of code. As far as I know (I could be wrong), that is not necessary, but I leave it to you whether there's a more efficient path than skipping to the affine code at that point.

I installed CVXQUAD following this link
http://ask.cvxr.com/t/cvxquad-how-to-use-cvxquads-pade-approximant-instead-of-cvxs-unreliable-successive-approximation-for-gp-mode-log-exp-entr-rel-entr-kl-div-log-det-det-rootn-exponential-cone-cvxquads-quantum-matrix-entropy-matrix-log-related-functions/5598
but received the error: Undefined function or variable 'op_rel_entr_epi_cone'.