[mlpack-svn] [mlpack] add a test case for max-cut (#372)

Ryan Curtin notifications at github.com
Sat Jan 3 01:04:46 EST 2015


The test itself looks great to me, but after checking out your branch and building the test I get the following output:

```
...
[WARN ] Penalty is 3.50223e-20 (threshold 1.93732e-20).
[WARN ] Updated sigma to 2e+06.
[WARN ] AugLagrangian on iteration 16, starting with objective -1.2748.
[DEBUG] L-BFGS iteration 0; objective -3672.74.
[DEBUG] Line search failed.  Terminating optimization.
/home/ryan/src/stephentu/mlpack/src/mlpack/tests/lrsdp_test.cpp(181): fatal error in "ErdosRenyiRandomGraphMaxCutSDP": difference{99.9653%} between finalValue{-1.274800020898166} and -3672.7{-3672.6999999999998} exceeds 0.10000000000000001%

*** 1 failure detected in test suite "MLPACKTest"
```

I'm actually on the road right now (figuratively, not literally) so I can't dig deeper into what's going on here, but the L-BFGS objective seems to be correct, but what's being returned by AugLagrangian isn't right.  I'll dig deeper if I have a chance in the next couple of days, but I did want to at least express that people don't often like to write test cases, so I really appreciate it.  I'm glad to see this LRSDP code finally be useful!  (Or, getting close to there, at least. :))

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/372#issuecomment-68584781
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20150102/5d50f437/attachment.html>


More information about the mlpack-git mailing list