New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Lovasz-Theta AugLagrangian tests do not work #182

Closed
rcurtin opened this Issue Dec 29, 2014 · 2 comments

Comments

Projects
None yet
1 participant
@rcurtin
Member

rcurtin commented Dec 29, 2014

Reported by rcurtin on 23 Aug 42063786 11:54 UTC
I am unable to replicate the results of Monteiro & Burer, 2004. This may mean that the Augmented Lagrangian algorithm itself has a bug in it (which I suspect), or simply that my implementation of the Lovasz-Theta SDP is incorrect (which I suspect slightly less).

When this is solved, tests should be working for:

  • johnson8-4-4
  • hamming6-4
  • keller4

Runtime may be too long on those, so maybe we'll only run one each time, but before this is resolved all those tests need to work.

@rcurtin rcurtin self-assigned this Dec 29, 2014

@rcurtin rcurtin added this to the mlpack 1.0.1 milestone Dec 29, 2014

@rcurtin rcurtin closed this Dec 29, 2014

@rcurtin

This comment has been minimized.

Show comment
Hide comment
@rcurtin

rcurtin Dec 30, 2014

Member

Commented by rcurtin on 26 Sep 42102660 00:43 UTC
The johnson8-4-4 test finally works. The other two still need to be done. It is also slow but I am working on that.

Member

rcurtin commented Dec 30, 2014

Commented by rcurtin on 26 Sep 42102660 00:43 UTC
The johnson8-4-4 test finally works. The other two still need to be done. It is also slow but I am working on that.

@rcurtin

This comment has been minimized.

Show comment
Hide comment
@rcurtin

rcurtin Dec 30, 2014

Member

Commented by rcurtin on 27 Apr 42104889 18:55 UTC
The hamming6-4 test did not converge (likely just bad initial Lagrange multipliers), but the keller4 test did. This is sufficient for me to say that LRSDP is working. However, the keller4 test is slow (~5min without debugging or profiling symbols), so I will leave it commented out.

Member

rcurtin commented Dec 30, 2014

Commented by rcurtin on 27 Apr 42104889 18:55 UTC
The hamming6-4 test did not converge (likely just bad initial Lagrange multipliers), but the keller4 test did. This is sufficient for me to say that LRSDP is working. However, the keller4 test is slow (~5min without debugging or profiling symbols), so I will leave it commented out.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment