--- Log opened Sun Jun 03 00:00:41 2012 | ||
KMcQuisten | Where was it getting caught up? | 00:01 |
---|---|---|
blackburn | hmmm something strange here.. | 00:05 |
blackburn | KMcQuisten: funny thing x2 factor comes from the code, but not two feature instances.. | 00:05 |
KMcQuisten | Really? I wonder if it's pulling them from the initialized kernels | 00:06 |
blackburn | yeah.. | 00:07 |
KMcQuisten | I tired a CombinedFeature and CombinedKernel with StringChar features and WD kernels of 2 different orders and got the same sort of error as with the StringWordFeatures | 00:12 |
KMcQuisten | SystemError: [ERROR] assertion char_vec && len>0 failed in file kernel/WeightedDegreeStringKernel.cpp line 522 | 00:12 |
blackburn | different error | 00:13 |
KMcQuisten | Yes, but still failing the check to see if the proper number of features are present | 00:13 |
KMcQuisten | The same thing happens when I combine a WD kernel on StringChar features with a CommWOrdString kernel on StringWord features. | 00:15 |
KMcQuisten | Like you, I hae stringChar+Real and StringWord + real working fine | 00:16 |
blackburn | heh | 00:17 |
KMcQuisten | Somethign about this doesn't like having multiple kinds of string features in a CombinedKernel | 00:17 |
KMcQuisten | Very strange | 00:17 |
blackburn | does it happen with real+real? | 00:17 |
KMcQuisten | Let me check quicklty | 00:18 |
blackburn | hmm it happens with real+string now | 00:20 |
KMcQuisten | Real and Real worked for me | 00:21 |
KMcQuisten | Real and StringWord work for me | 00:21 |
KMcQuisten | Let me check Real and StringChar | 00:21 |
KMcQuisten | Real and StringChar work for me with a WD kernel | 00:24 |
KMcQuisten | ONly thing that fials for me is combining multibple string features | 00:24 |
KMcQuisten | man my typing is crap today. | 00:25 |
blackburn | heh | 00:25 |
KMcQuisten | Need more sleep :) | 00:25 |
blackburn | no problem | 00:25 |
blackburn | okay but real + string fails now too | 00:25 |
blackburn | real+real works | 00:27 |
KMcQuisten | Not for me. Real + string still works for me | 00:28 |
KMcQuisten | real + real works for me, too | 00:28 |
KMcQuisten | I set it up real + string and string + real. Both work | 00:28 |
blackburn | ok more or less it is clear now | 00:29 |
blackburn | I'll dig into the code again | 00:29 |
KMcQuisten | what's the verdict? | 00:29 |
KMcQuisten | Allrighty | 00:29 |
KMcQuisten | Ok, i'm going ot head out of here. When you have news feel free to let me know on the mailing list. | 00:33 |
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Quit: Page closed] | 00:33 | |
blackburn | ok | 00:40 |
CIA-9 | shogun: Sergey Lisitsyn master * r4b8c7d0 / (2 files in 2 dirs): Hopefully fixed MKL linadd in SVRLight - http://git.io/uNj-bQ | 01:18 |
shogun-buildbot | build #576 of ruby_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/576 blamelist: blackburn91@gmail.com | 02:36 |
shogun-buildbot | build #580 of python_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/580 blamelist: blackburn91@gmail.com | 03:02 |
CIA-9 | shogun: Sergey Lisitsyn master * r17a1306 / examples/undocumented/python_modular/mkl_linadd_regression.py : Commented MKL breaker in example to make test work - http://git.io/cnlkBw | 03:24 |
-!- blackburn [~blackburn@31.28.59.65] has quit [Ping timeout: 244 seconds] | 03:28 | |
shogun-buildbot | build #577 of ruby_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/ruby_modular/builds/577 | 04:36 |
shogun-buildbot | build #581 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_modular/builds/581 | 04:52 |
-!- pluskid [~pluskid@111.120.67.183] has joined #shogun | 07:55 | |
-!- pluskid [~pluskid@111.120.67.183] has quit [Ping timeout: 252 seconds] | 09:30 | |
-!- pluskid [~pluskid@li400-235.members.linode.com] has joined #shogun | 09:30 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 10:06 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 10:28 | |
-!- blackburn [~blackburn@31.28.59.65] has joined #shogun | 11:24 | |
pluskid | blackburn: do you know how to clone a VowpalWabbitMachine? | 11:29 |
blackburn | a=b.clone()? | 11:29 |
pluskid | is there such a method? | 11:29 |
blackburn | I don't think so | 11:29 |
pluskid | hmm, too bad | 11:30 |
pluskid | I want to have a copy of a VWMachine somewhere | 11:30 |
blackburn | add a copy constructor :) | 11:30 |
pluskid | seems I have to construct the copy manually | 11:30 |
pluskid | to do that then I should have to understand the implementation of VW first | 11:30 |
pluskid | btw: since shogun has a generic serialization component, I guess it should not be too difficult to build a generic clone component based on that | 11:31 |
blackburn | virtual clone method? | 11:32 |
pluskid | I'm not sure, but I guess it could be implemented in a similar way to serialization | 11:34 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 11:34 | |
pluskid | for example, a *very naive* way is to serialize it to some where and then de-serialize, we get a clone ... | 11:34 |
blackburn | I don't mind actually if somebody do that :) | 11:36 |
pluskid | :D | 11:37 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 12:36 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 244 seconds] | 13:43 | |
-!- blackburn [~blackburn@31.28.59.65] has left #shogun [] | 14:47 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 15:02 | |
-!- pluskid [~pluskid@li400-235.members.linode.com] has quit [Quit: Leaving] | 15:59 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 18:13 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 18:13 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 18:13 | |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has joined #shogun | 18:47 | |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has quit [Ping timeout: 245 seconds] | 19:42 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds] | 19:51 | |
-!- wiking_ [~wiking@208.76.55.198] has joined #shogun | 20:11 | |
-!- wiking_ [~wiking@208.76.55.198] has quit [Changing host] | 20:11 | |
-!- wiking_ [~wiking@huwico/staff/wiking] has joined #shogun | 20:11 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 252 seconds] | 20:12 | |
-!- wiking_ is now known as wiking | 20:12 | |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has joined #shogun | 20:12 | |
CIA-9 | shogun: Soeren Sonnenburg master * r7fb034c / (4 files in 2 dirs): add liblinear's L2R_LR - http://git.io/7cx9Ew | 21:04 |
-!- heiko [~heiko@host86-185-9-122.range86-185.btcentralplus.com] has joined #shogun | 21:07 | |
-!- heiko [~heiko@host86-185-9-122.range86-185.btcentralplus.com] has quit [Ping timeout: 260 seconds] | 21:23 | |
-!- heiko [~heiko@host86-183-73-53.range86-183.btcentralplus.com] has joined #shogun | 21:36 | |
-!- heiko [~heiko@host86-183-73-53.range86-183.btcentralplus.com] has quit [Ping timeout: 260 seconds] | 21:45 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 22:05 | |
CIA-9 | shogun: Soeren Sonnenburg master * r9b380e8 / (7 files in 5 dirs): Integrate liblinears SVR - http://git.io/n9UVCg | 22:36 |
n4nd0 | hey wiking | 22:49 |
n4nd0 | how did it go in the conference? | 22:50 |
n4nd0 | still there? | 22:50 |
wiking | arrived today | 22:50 |
wiking | i mean i came back today | 22:50 |
n4nd0 | good | 22:50 |
n4nd0 | did you enjoy it? | 22:50 |
-!- blackburn [~blackburn@31.28.59.65] has joined #shogun | 22:55 | |
blackburn | wiking: heya | 22:55 |
wiking | yeah it was great | 22:55 |
wiking | but today i still need to recover | 22:56 |
wiking | blackburn: yo | 22:56 |
blackburn | when do you think you can get back on the track? | 22:57 |
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has joined #shogun | 23:00 | |
KMcQuisten | Hello all | 23:00 |
blackburn | hey | 23:00 |
KMcQuisten | Get my email? | 23:00 |
blackburn | which one? | 23:00 |
KMcQuisten | most recent. sent today | 23:01 |
KMcQuisten | moments ago, inf act | 23:01 |
blackburn | ah right I just sent an answer | 23:01 |
blackburn | 10 mins ago I guess | 23:01 |
KMcQuisten | well then let me look | 23:01 |
blackburn | ok I suggested to turn off linadd at all | 23:01 |
KMcQuisten | ok let me check | 23:02 |
-!- heiko [~heiko@109.144.207.254] has joined #shogun | 23:05 | |
blackburn | it can be that linadd is a little broken | 23:06 |
KMcQuisten | Looks like tha'ts the case. It works with LINADD disables | 23:07 |
KMcQuisten | d | 23:07 |
wiking | blackburn: tomorrow | 23:10 |
blackburn | cool | 23:12 |
blackburn | KMcQuisten: cool too :) | 23:12 |
KMcQuisten | Looks like we got this squared away for now. | 23:12 |
KMcQuisten | Thanks for all your helop | 23:12 |
blackburn | no problem! ok then I know something is broken in linadd here | 23:12 |
KMcQuisten | Glad I could help root these out. | 23:13 |
KMcQuisten | I think I may have found another one, though. | 23:14 |
KMcQuisten | if I do mkl.get_kernel().get_subkernel_weights() to extract the weights after training, I get this: | 23:14 |
KMcQuisten | Traceback (most recent call last): File "mkl_linadd_regression.py", line 76, in <module> mkl_binclass_modular (*parameter_list[0]) File "mkl_linadd_regression.py", line 71, in mkl_binclass_modular print mkl.get_kernel().get_subkernel_weights() | 23:16 |
KMcQuisten | TypeError: Kernel_get_subkernel_weights expected 1 arguments, got 0 | 23:16 |
blackburn | hmm | 23:17 |
KMcQuisten | If I just use the method on a frech CombinedKernel, everything works fine, but if I pass a kernel to a machine and train it, I get that error when I try to extract the weights | 23:17 |
blackburn | ahh I think I know | 23:17 |
KMcQuisten | righto | 23:18 |
blackburn | different types | 23:18 |
KMcQuisten | Yes. when I extract the kernel, it's just type "Kernel", | 23:18 |
blackburn | get_kernel() returns CKernel | 23:18 |
blackburn | so rather use old pointer or cast it somehow | 23:18 |
KMcQuisten | Then how should I get my subkernel weights after I train an MKL? | 23:18 |
blackburn | let me check why CKernel has that method | 23:18 |
blackburn | well you have combined kernel somewhere in code | 23:19 |
blackburn | just call get_subkernel_weights on that | 23:19 |
blackburn | it will be the same kernel still after train too | 23:19 |
KMcQuisten | if I call the method from the variable I used to set up the combined kernel, it segfaults | 23:19 |
blackburn | oh | 23:19 |
blackburn | cool | 23:19 |
blackburn | let me fix it then | 23:19 |
-!- heiko [~heiko@109.144.207.254] has quit [Ping timeout: 252 seconds] | 23:20 | |
blackburn | should be easy | 23:20 |
-!- heiko [~heiko@host86-181-155-21.range86-181.btcentralplus.com] has joined #shogun | 23:27 | |
CIA-9 | shogun: Heiko Strathmann master * r4343cf8 / (2 files): some initial work on Gram-Eigen-Spectrum based Threshold for MMD^2 - http://git.io/h73zWg | 23:27 |
CIA-9 | shogun: Heiko Strathmann master * re04a0e6 / (3 files in 2 dirs): -added convienience methods for display_matrix/vector based on SGVector - http://git.io/wmDC7Q | 23:27 |
CIA-9 | shogun: Heiko Strathmann master * r96dc418 / (2 files): -added create_identity_matrix based on SGMatrix - http://git.io/trXx2A | 23:27 |
CIA-9 | shogun: Heiko Strathmann master * r525eb66 / examples/undocumented/libshogun/mathematics_lapack.cpp : usage of create_identity_matrix - http://git.io/ekx0Ow | 23:27 |
CIA-9 | shogun: Heiko Strathmann master * rf91331a / (5 files in 3 dirs): Merge pull request #563 from karlnapf/master - http://git.io/mMTKXw | 23:27 |
blackburn | KMcQuisten: I'll commit in a min | 23:28 |
blackburn | or two or three :D | 23:28 |
KMcQuisten | Righto | 23:28 |
blackburn | argh | 23:37 |
KMcQuisten | ? | 23:37 |
blackburn | fixed in my fork instead of main repo :D | 23:37 |
KMcQuisten | Ha! Whoops | 23:37 |
blackburn | KMcQuisten: done - update your repo and recompile - should work | 23:39 |
blackburn | argh | 23:39 |
blackburn | not yet | 23:39 |
KMcQuisten | ok | 23:39 |
KMcQuisten | I'll wiat | 23:39 |
KMcQuisten | :) | 23:39 |
blackburn | now yes | 23:39 |
CIA-9 | shogun: Sergey Lisitsyn master * r71ddf75 / src/shogun/kernel/CombinedKernel.cpp : Fixed get subkernel weights in CombinedKernel - http://git.io/cGIC9w | 23:40 |
KMcQuisten | Works fine. | 23:48 |
KMcQuisten | Thanks | 23:48 |
blackburn | welcome! | 23:50 |
KMcQuisten | I"ll definitely let you know if anything else pops up. | 23:51 |
blackburn | heh ok | 23:51 |
KMcQuisten | Thanks so much , again | 23:51 |
-!- KMcQuisten [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Quit: Page closed] | 23:51 | |
--- Log closed Mon Jun 04 00:00:41 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!