--- Log opened Fri Jul 19 00:00:37 2013 | ||
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 00:02 | |
-!- foulwall` [~user@dirtycod.es] has quit [Ping timeout: 240 seconds] | 00:10 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 00:20 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 00:20 | |
-!- nube [~rho@49.244.52.224] has joined #shogun | 01:02 | |
-!- zxtx [~zv@rrcs-76-79-81-162.west.biz.rr.com] has joined #shogun | 01:03 | |
shogun-notifier- | shogun: Sergey Lisitsyn :develop * e95673c / / (3 files): https://github.com/shogun-toolbox/shogun/commit/e95673cc678c6ad9626b29010369a34885f4ff12 | 01:09 |
---|---|---|
shogun-notifier- | shogun: Templated classes clone unit-testing | 01:09 |
shogun-buildbot | build #1475 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1475 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 01:26 |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has joined #shogun | 01:34 | |
travis-ci | [travis-ci] it's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9247772 | 01:34 |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has left #shogun [] | 01:34 | |
@iglesiasg | good night! | 02:26 |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 02:27 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Quit: Leaving] | 02:37 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 02:37 | |
-!- zxtx [~zv@rrcs-76-79-81-162.west.biz.rr.com] has quit [Ping timeout: 248 seconds] | 02:41 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Quit: Leaving] | 02:50 | |
shogun-buildbot | build #397 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/397 | 03:08 |
-!- nube [~rho@49.244.52.224] has quit [Ping timeout: 240 seconds] | 03:47 | |
-!- nube [~rho@49.244.92.33] has joined #shogun | 04:01 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 04:09 | |
shogun-buildbot | build #462 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/462 | 04:19 |
-!- nube [~rho@49.244.92.33] has quit [Quit: Leaving.] | 04:58 | |
-!- nube [~rho@116.90.239.13] has joined #shogun | 05:47 | |
-!- splatvector [617fb9c0@gateway/web/freenode/ip.97.127.185.192] has joined #shogun | 06:27 | |
splatvector | Hello | 06:28 |
splatvector | anyone awake? | 06:33 |
-!- splatvector [617fb9c0@gateway/web/freenode/ip.97.127.185.192] has quit [Quit: Page closed] | 06:35 | |
-!- foulwall` [~user@2001:da8:215:503:65ec:4c22:aeb8:84e7] has joined #shogun | 07:16 | |
-!- heru [~heru@182.8.122.99] has joined #shogun | 07:20 | |
-!- heru [~heru@182.8.122.99] has quit [Quit: -a-] | 08:03 | |
-!- gsomix_ [~gsomix@109.169.233.216] has joined #shogun | 08:43 | |
-!- gsomix [~gsomix@178.45.92.6] has quit [Ping timeout: 268 seconds] | 08:46 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 08:57 | |
-!- nube [~rho@116.90.239.13] has quit [Quit: Leaving.] | 09:09 | |
-!- foulwall` [~user@2001:da8:215:503:65ec:4c22:aeb8:84e7] has quit [Remote host closed the connection] | 09:09 | |
-!- foulwall [~user@2001:da8:215:503:34e9:d67c:516d:eee3] has joined #shogun | 09:16 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 09:31 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 09:31 | |
@iglesiasg | good morning guys | 09:32 |
gsomix_ | iglesiasg, hey | 09:37 |
gsomix_ | iglesiasg, I'm alive. Insects just took my money. | 09:39 |
@iglesiasg | gsomix_: haha, Russian insects! | 09:39 |
-!- foulwall [~user@2001:da8:215:503:34e9:d67c:516d:eee3] has quit [Remote host closed the connection] | 10:50 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 10:57 | |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 240 seconds] | 11:01 | |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has joined #shogun | 11:13 | |
-!- nube [~rho@116.90.239.13] has joined #shogun | 11:19 | |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has joined #shogun | 11:25 | |
-!- gsomix_ is now known as gsomix | 11:34 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Read error: Operation timed out] | 11:34 | |
-!- HeikoS [~heiko@nat-177-148.internal.eduroam.ucl.ac.uk] has joined #shogun | 11:44 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 11:44 | |
@HeikoS | sonney2k, lisitsyn, wiking, please have a look at the new issues I put on github, let's discuss those in the thread a bit to have the ideas documented. Curious about your opinions | 11:47 |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 12:14 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 12:14 | |
-!- nube [~rho@116.90.239.13] has quit [Quit: Leaving.] | 12:15 | |
van51 | gsomix: hey | 12:18 |
gsomix | van51, sup? | 12:18 |
van51 | gsomix: I saw you talking with sonney yesterday but I didn't read the entire conversation | 12:19 |
van51 | gsomix: do you still need me to modify the tokenizer? | 12:19 |
gsomix | van51, nope. | 12:19 |
van51 | gsomix: ok then | 12:20 |
van51 | gsomix: one last question | 12:20 |
van51 | gsomix: what other parsers are you going to do? | 12:20 |
gsomix | van51, libsvm, protobuf, matlab/octave at least. | 12:22 |
gsomix | but now I'm little stuck with initial stuff. so we'll see. | 12:23 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 12:25 | |
shogun-notifier- | shogun: Kevin :develop * 8663f33 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/8663f33ffa8261a217931b9bab6b3c97bf5eda52 | 12:25 |
shogun-notifier- | shogun: added uwedge approximate joint diagonalizer and unit test | 12:25 |
shogun-notifier- | shogun: Kevin :develop * 3ba868d / src/shogun/mathematics/ajd/ (3 files): https://github.com/shogun-toolbox/shogun/commit/3ba868db83c06fdb932df85244e80f78f9b97632 | 12:25 |
shogun-notifier- | shogun: added citations for the ajd algorithms | 12:25 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 9985650 / / (5 files): https://github.com/shogun-toolbox/shogun/commit/99856506cbf55f48d73ef43192031aa627f3c1b9 | 12:25 |
shogun-notifier- | shogun: Merge pull request #1255 from pickle27/develop | 12:25 |
shogun-notifier- | shogun: | 12:25 |
shogun-notifier- | shogun: added uwedge approximate joint diagonalizer and unit test | 12:25 |
van51 | gsomix: is libsvm format the same as svmlight? | 12:25 |
van51 | gsomix: I'm asking because I found a read_svmlight_features and I thought maybe you could use it as base or guide, although it's a bit messy | 12:27 |
van51 | method* | 12:27 |
van51 | gsomix: or if you did something new, we could update that | 12:28 |
gsomix | van51, yep, it's the same. | 12:29 |
gsomix | van51, yeah, I plan update some old stuff. | 12:29 |
van51 | gsomix: ok then | 12:30 |
van51 | gsomix: that was all :) | 12:30 |
shogun-buildbot | build #1477 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1477 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>, Kevin <kevinhughes27@gmail.com> | 12:34 |
thoralf | Hey guys. | 12:35 |
@iglesiasg | Hi thoralf | 12:36 |
thoralf | HeikoS: I've thought about the test generator. What about using GCCXML to parse the header files, extract all possible constructors and indeed run automatic tests if we can handle the parameters. | 12:38 |
thoralf | HeikoS: For example we could handle int32_t, SGMatrix, etc. | 12:38 |
@HeikoS | thoralf: I also had another idea yesterday night | 12:39 |
@HeikoS | thoralf: since some classes need special types of input | 12:39 |
@HeikoS | i.e. not just a matrix but a psd matrix | 12:39 |
@HeikoS | this is very hard to handle automatically | 12:39 |
@HeikoS | but we could do a hybrid approach | 12:39 |
thoralf | psd? | 12:39 |
@HeikoS | positive matrix | 12:39 |
thoralf | Ah. | 12:39 |
@HeikoS | say CSGobject gets an abstract method that creates a non-trivial instance | 12:40 |
@HeikoS | then developers are forced to create this | 12:40 |
@HeikoS | which makes sense since they know how a non-trivial instance looks like | 12:40 |
@HeikoS | example: | 12:40 |
@HeikoS | CSGObject* get_non_trivial_instance() | 12:41 |
@HeikoS | for dense features | 12:41 |
thoralf | HeikoS: This limits us to one instance. | 12:41 |
@HeikoS | this could just create a random matrix, add a subset and a preprocessor and return | 12:41 |
@HeikoS | thoralf: yes, but the instance makes sense since its not generated automatically | 12:41 |
@HeikoS | thoralf: we can easily return a list and if people want to test more, they can do that | 12:41 |
@HeikoS | thoralf: this way we can have all the tests that can be automated, serialisation, equals/clone | 12:42 |
@HeikoS | but working on non-trivial instances | 12:42 |
thoralf | HeikoS: Why so complicated? ;) | 12:42 |
@HeikoS | while these are handwritten | 12:42 |
thoralf | HeikoS: I won't change existing classes for this. | 12:42 |
@HeikoS | thoralf: yes thats the downside | 12:42 |
thoralf | HeikoS: We're messing up design for automatic test generator. | 12:42 |
@HeikoS | thoralf: but since our classes are quite diverse I dont know about automagic tests | 12:42 |
@HeikoS | there are constraints on the data you can give to a class | 12:43 |
@HeikoS | and what actually appears in practice | 12:43 |
@HeikoS | if we do this via types we might run into problems | 12:43 |
@HeikoS | thoralf: oh and btw we can stack this get_non_trivial instance | 12:43 |
@HeikoS | for example an svm can simply call DenseFeatures::get_non_trivial instance | 12:43 |
@HeikoS | thoralf: but I am open for other approaches | 12:44 |
@HeikoS | thoralf: maybe good to discuss this also with lisitsyn and wiking | 12:44 |
@HeikoS | thoralf: you can also start a discussion on github issues, we havve a new label: "things to discuss" | 12:44 |
-!- gsomix [~gsomix@109.169.233.216] has quit [Remote host closed the connection] | 12:44 | |
@HeikoS | thoralf: gotta do some work now, be back later today :) | 12:44 |
thoralf | HeikoS: One second please. ;) | 12:45 |
@HeikoS | thoralf: ok :) | 12:45 |
thoralf | HeikoS: My idea would be a simple class for generating examples. For example several SGMatrix instances. Each provided by a simple method "generator_1x1_matrix()", etc. We then can simply use this for automagic code generation. All we need to know is: Possible classes+their constructors *and* possible data generators. | 12:46 |
shogun-buildbot | build #1476 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1476 | 12:46 |
thoralf | HeikoS: Both can be found with gccxml and be parsed with python. | 12:47 |
@HeikoS | thoralf: SGmatrix cannot be tested automatically btw | 12:47 |
@HeikoS | how would that work for example for an SVM with all its parameters | 12:47 |
@HeikoS | we then have to define those methods for every class, right? | 12:47 |
thoralf | HeikoS: Yes and no. | 12:47 |
shogun-buildbot | build #1057 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1057 blamelist: Kevin <kevinhughes27@gmail.com> | 12:48 |
shogun-buildbot | build #743 of rpm1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/743 blamelist: Kevin <kevinhughes27@gmail.com> | 12:48 |
thoralf | If we find a class LibLinear(int, dotfeatures, labels), we can just insert the example data for each parameter, | 12:49 |
@HeikoS | thoralf: so we have to define all types anyway | 12:50 |
@HeikoS | thoralf: why then not do this within a class, much less hacking and effectively the same | 12:50 |
@HeikoS | thoralf: and actually modular since one can stack things | 12:50 |
thoralf | HeikoS: We only define the types we want to support. | 12:50 |
@HeikoS | thoralf: all CSGObjects | 12:51 |
thoralf | HeikoS: I think the tests are not part of shogun, so keep them out of the classes. | 12:51 |
@HeikoS | thoralf: thats true! | 12:51 |
thoralf | HeikoS: They're just QA code. | 12:51 |
@HeikoS | thoralf: maybe define this in another file then | 12:52 |
thoralf | HeikoS: But what about OO? Just overloading? | 12:52 |
@HeikoS | thoralf: yes | 12:52 |
thoralf | HeikoS: SGMatrixTestClass1 ;) | 12:52 |
thoralf | extends SGMatrix | 12:52 |
@HeikoS | thoralf: no one can add methods in other files | 12:52 |
@HeikoS | SGMatrix cannot be tested this way anyway | 12:52 |
@HeikoS | only SGObjet instances | 12:52 |
thoralf | Oh. | 12:53 |
thoralf | Yes. | 12:53 |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has joined #shogun | 12:53 | |
travis-ci | [travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9261914 | 12:53 |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has left #shogun [] | 12:53 | |
thoralf | HeikoS: Having data generators would help for testing, anyway. Maybe we could isolate them (no matter if they're factories, subclasses, etc.) | 12:54 |
@HeikoS | thoralf: yes that would be good indeed | 12:54 |
thoralf | HeikoS: Thanks, I think I made my point. | 12:54 |
@HeikoS | thoralf: could you add a thread on giuthub on this? | 12:54 |
thoralf | HeikoS: I was about to ask you that. ;) | 12:54 |
thoralf | HeikoS: It was your issue. ;) | 12:54 |
@HeikoS | thoralf: well I can do it too, then you can add our idea and we can discuss with the others | 12:55 |
shogun-buildbot | build #1261 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1261 blamelist: Kevin <kevinhughes27@gmail.com> | 12:57 |
@iglesiasg | HeikoS: Hi! Any idea what is it with the clone method in unit tests? | 13:00 |
@HeikoS | iglesiasg: it needs some work :) | 13:00 |
@HeikoS | iglesiasg: just added and reveals many bugs | 13:00 |
@HeikoS | working on it | 13:00 |
@HeikoS | dont worry about them now | 13:00 |
@iglesiasg | HeikoS: ok! thank you :) | 13:00 |
@HeikoS | in the future, this will detect many errors when one creates new classes | 13:00 |
@iglesiasg | that sounds like a good thing | 13:01 |
-!- gsomix [~gsomix@109.169.233.216] has joined #shogun | 13:22 | |
-!- foulwall [~user@2001:da8:215:503:34e9:d67c:516d:eee3] has joined #shogun | 13:25 | |
shogun-buildbot | build #1380 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1380 blamelist: Kevin <kevinhughes27@gmail.com> | 13:28 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has quit [Ping timeout: 264 seconds] | 13:32 | |
-!- FSCV [~FSCV@50.7.50.60] has joined #shogun | 13:56 | |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has joined #shogun | 14:08 | |
-!- foulwall` [~user@2001:da8:215:503:6c94:7e66:a314:1a03] has joined #shogun | 14:14 | |
lambday | HeikoS: hi | 14:15 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has quit [Quit: Leaving.] | 14:15 | |
-!- foulwall [~user@2001:da8:215:503:34e9:d67c:516d:eee3] has quit [Ping timeout: 245 seconds] | 14:16 | |
-!- foulwall` [~user@2001:da8:215:503:6c94:7e66:a314:1a03] has quit [Remote host closed the connection] | 14:20 | |
-!- foulwall` [~user@2001:da8:215:503:6c94:7e66:a314:1a03] has joined #shogun | 14:20 | |
-!- foulwall` [~user@2001:da8:215:503:6c94:7e66:a314:1a03] has quit [Ping timeout: 264 seconds] | 14:33 | |
lambday | HeikoS: I have added COCG solver.. (this one we'll be needing for individual solve since complex shifts will be in the diagonal).. | 14:35 |
lambday | HeikoS: the clone msgs are really helpful :) | 14:35 |
lambday | HeikoS: currently github is down.. I'll send a PR as soon as its up.. | 14:35 |
lambday | HeikoS: I'll add COCG_M after this | 14:36 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has quit [] | 14:48 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 15:08 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 15:25 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 15:56 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 15:56 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 16:24 | |
shogun-notifier- | shogun: van51 :develop * d8855dc / src/shogun/features/HashedDocDotFeatures.cpp,benchmarks/hasheddoc_benchmarks.cpp: https://github.com/shogun-toolbox/shogun/commit/d8855dc3fffae31c2d9d005d35cc7b4ffd86cae1 | 16:24 |
shogun-notifier- | shogun: Changed norm constant in HashedDocDotFeatures | 16:24 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 646d0a6 / src/shogun/features/HashedDocDotFeatures.cpp,benchmarks/hasheddoc_benchmarks.cpp: https://github.com/shogun-toolbox/shogun/commit/646d0a6d74e416156acf24a5e56f7612ee6b8a42 | 16:24 |
shogun-notifier- | shogun: Merge pull request #1266 from van51/feature/hashing | 16:24 |
shogun-notifier- | shogun: | 16:24 |
shogun-notifier- | shogun: Changed norm constant in HashedDocDotFeatures | 16:24 |
@sonney2k | van51, if you have time to give them an overhaul I don't mind. I am not particularly excited about VW's format but if it would work with the other shogun classes it is useful! | 16:25 |
shogun-notifier- | shogun: van51 :develop * 9268bff / / (6 files): https://github.com/shogun-toolbox/shogun/commit/9268bff1ef981290fd4e119ec4f0bdacb858680b | 16:26 |
shogun-notifier- | shogun: StreamingHashedSparseFeatures class | 16:26 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 42fbf1b / / (6 files): https://github.com/shogun-toolbox/shogun/commit/42fbf1bd61ff40b84d225c206e696d8450873e31 | 16:26 |
shogun-notifier- | shogun: Merge pull request #1262 from van51/feature/streaming_sparse | 16:26 |
shogun-notifier- | shogun: | 16:26 |
shogun-notifier- | shogun: StreamingHashedSparseFeatures class | 16:26 |
shogun-buildbot | build #1479 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1479 blamelist: Soeren Sonnenburg <sonne@debian.org> | 16:28 |
@sonney2k | HeikoS, wiking any chance to get the build tests back to green | 16:28 |
@sonney2k | we will miss out on other bugs if we don't do that soonish | 16:29 |
gsomix | sonney2k, hey. can I use DynArray's for store lines that I read from csv? | 16:31 |
gsomix | there is needed for matrices reading. | 16:31 |
shogun-buildbot | build #1480 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1480 blamelist: van51 <vangelis_51@hotmail.com> | 16:33 |
shogun-buildbot | build #1481 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1481 blamelist: Soeren Sonnenburg <sonne@debian.org> | 16:36 |
shogun-buildbot | build #1478 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1478 blamelist: van51 <vangelis_51@hotmail.com> | 16:41 |
-!- nube [~rho@49.244.49.217] has joined #shogun | 16:46 | |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has joined #shogun | 16:58 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9269142 | 16:58 |
-!- travis-ci [~travis-ci@ec2-107-22-157-209.compute-1.amazonaws.com] has left #shogun [] | 16:58 | |
-!- nube [~rho@49.244.49.217] has quit [Quit: Leaving.] | 17:00 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 17:02 | |
-!- foulwall` [~user@2001:da8:215:c252:20cc:92aa:1f23:510c] has joined #shogun | 17:04 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 17:11 | |
-!- nube [~rho@49.244.49.217] has joined #shogun | 17:22 | |
pickle27 | lisitsyn: hey | 17:24 |
-!- travis-ci [~travis-ci@ec2-54-226-190-137.compute-1.amazonaws.com] has joined #shogun | 17:24 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9269250 | 17:24 |
-!- travis-ci [~travis-ci@ec2-54-226-190-137.compute-1.amazonaws.com] has left #shogun [] | 17:24 | |
pickle27 | or sonney2k here? | 17:24 |
-!- nube [~rho@49.244.49.217] has quit [Client Quit] | 17:25 | |
@HeikoS | sonney2k: yes, waiting for some patches for that by lisitsyn | 17:34 |
shogun-notifier- | shogun: lambday :develop * 6936e72 / / (10 files): https://github.com/shogun-toolbox/shogun/commit/6936e720a5d201b61308228737d795b672fed860 | 17:37 |
shogun-notifier- | shogun: cocg solver added, few more changes made in log-det | 17:37 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 938cb02 / / (10 files): https://github.com/shogun-toolbox/shogun/commit/938cb02c25ad8f57ffd7003d950b56c5c26fdc52 | 17:37 |
shogun-notifier- | shogun: Merge pull request #1267 from lambday/feature/log_determinant | 17:37 |
shogun-notifier- | shogun: | 17:37 |
shogun-notifier- | shogun: COCG solver added in log-det framework | 17:37 |
pickle27 | hey guys | 17:40 |
pickle27 | I have a question | 17:40 |
pickle27 | is there a way with shogun to test if an SGMatrix is positive definite? | 17:40 |
-!- FSCV [~FSCV@50.7.50.60] has quit [Quit: Leaving] | 17:41 | |
pickle27 | and if there isn't I'd like to add one, should this be a member of SGMatrix or a function in one of the modules | 17:41 |
-!- foulwall` [~user@2001:da8:215:c252:20cc:92aa:1f23:510c] has quit [Remote host closed the connection] | 17:45 | |
-!- nube [~rho@49.244.49.217] has joined #shogun | 17:49 | |
shogun-buildbot | build #1483 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1483 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 17:51 |
thoralf | What happens, if I create an object on the stack and pass a reference inside a method which calls SG_REF/SG_UNREF? | 17:53 |
thoralf | Can shoguns reference counting handle this? | 17:54 |
thoralf | I'm getting strange errors at the end of the scope. | 17:55 |
shogun-buildbot | build #1482 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1482 blamelist: lambday <heavensdevil6909@gmail.com> | 17:56 |
thoralf | A minimal example: { CBinaryLabels y(100); CLibLinear svm = CLibLinear(); svm.set_labels(&y); } | 17:58 |
thoralf | Maybe someone can help: https://github.com/shogun-toolbox/shogun/issues/1268 | 18:06 |
thoralf | See you soon. | 18:06 |
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!] | 18:06 | |
-!- travis-ci [~travis-ci@ec2-54-226-190-137.compute-1.amazonaws.com] has joined #shogun | 18:15 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/9272060 | 18:15 |
-!- travis-ci [~travis-ci@ec2-54-226-190-137.compute-1.amazonaws.com] has left #shogun [] | 18:15 | |
-!- foulwall [~user@2001:da8:215:c252:c8ab:c19f:d597:37e7] has joined #shogun | 18:16 | |
-!- foulwall [~user@2001:da8:215:c252:c8ab:c19f:d597:37e7] has quit [Ping timeout: 245 seconds] | 18:21 | |
-!- nube [~rho@49.244.49.217] has quit [Quit: Leaving.] | 18:46 | |
-!- lambday [67157e36@gateway/web/freenode/ip.103.21.126.54] has joined #shogun | 18:57 | |
lambday | HeikoS: hi | 19:00 |
@HeikoS | lambday: hey man | 19:00 |
@HeikoS | how are things? | 19:00 |
-!- HeikoS [~heiko@nat-177-148.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 19:00 | |
lambday | ummm | 19:00 |
-!- HeikoS [~heiko@nat-177-148.internal.eduroam.ucl.ac.uk] has joined #shogun | 19:01 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:01 | |
lambday | HeikoS: hehe :D | 19:01 |
lambday | HeikoS: umm things are good | 19:01 |
lambday | HeikoS: will add COCG_M | 19:01 |
lambday | HeikoS: one thing about the eigensolver | 19:01 |
@HeikoS | lambday: yes? | 19:01 |
lambday | HeikoS: what shall we use for the tridigonal solver for Lanczos? | 19:02 |
lambday | tridiagonal* | 19:02 |
lambday | krylstat uses alglib | 19:02 |
@HeikoS | lambday: alglib has a tringular solver? | 19:02 |
@HeikoS | is it for sparse or dense things? | 19:02 |
@HeikoS | doesnt eigen have a triangular solver? | 19:03 |
@HeikoS | I remember erlend mentioning this was tricky to do with eigen | 19:03 |
lambday | I am not sure about whether eigen3 has it or not | 19:03 |
lambday | there are several implementations of lanczos | 19:04 |
lambday | one of them I found from the link that lisitsyn gave the other day | 19:04 |
lambday | HeikoS: https://github.com/lambday/KRYLSTAT/blob/master/lanczos/eigen/eigen_lanczos.h#L143 | 19:05 |
lambday | this is where they use this function smatrixtdevd | 19:06 |
lambday | that's in alglib | 19:06 |
lambday | I think it uses lapack inside it, since we already had lapack support, I can study what it does inside | 19:06 |
lambday | :-/ | 19:06 |
@HeikoS | we are talking about a triangular solver right? | 19:08 |
@HeikoS | isnt that easy? | 19:08 |
@HeikoS | just sustitide back in? | 19:08 |
@HeikoS | there must be a lib for this somewhere ;) | 19:08 |
@HeikoS | maybe have a look into alglib | 19:08 |
@HeikoS | we are using that one quite a bit already | 19:08 |
@HeikoS | so no problem | 19:09 |
@HeikoS | porting the functions to shogun is a bit painful though | 19:09 |
@HeikoS | but alglib is very mature | 19:09 |
lambday | HeikoS: we do have alglib support? | 19:12 |
lambday | that's awesome then... | 19:12 |
@HeikoS | lambday: we copy/paste code from algib | 19:12 |
@HeikoS | supporting it is impossible | 19:12 |
lambday | ohh :D | 19:12 |
@HeikoS | but we do proper porting of selected parts | 19:12 |
@HeikoS | have a look into CStatistics | 19:12 |
@HeikoS | many things in there use alglib | 19:12 |
@HeikoS | CDF functions, GAMMA etc | 19:12 |
@HeikoS | integrals | 19:12 |
lambday | okay I'll check | 19:13 |
lambday | if that doesn't work, will write from scratch | 19:13 |
@HeikoS | lambday: no use the algli one | 19:14 |
@HeikoS | you just have to replace some functions | 19:14 |
@HeikoS | like abs,exp,sin etc | 19:14 |
@HeikoS | you will see it | 19:14 |
@HeikoS | study an existing function before you do this, then you know what do to | 19:14 |
lambday | okay I'm checking | 19:14 |
@lisitsyn | HeikoS: re! | 19:38 |
@lisitsyn | what's up? | 19:38 |
@HeikoS | lisitsyn: what about the templates for the tests? | 19:41 |
@HeikoS | lisitsyn: sorry gotta go soon | 19:42 |
@lisitsyn | HeikoS: but they are here no? | 19:42 |
@HeikoS | lisitsyn: ah did not see :) | 19:43 |
@HeikoS | sorry busy day | 19:43 |
@lisitsyn | HeikoS: hehe | 19:43 |
@lisitsyn | HeikoS: the problem is that they fail | 19:43 |
@lisitsyn | I'll spend some time starting from now to check | 19:43 |
-!- nube [~rho@49.244.67.162] has joined #shogun | 19:44 | |
@HeikoS | arxiv | 19:44 |
@lisitsyn | arxiv what ;) | 19:46 |
@HeikoS | wrong window :) | 19:47 |
@lisitsyn | pickle27: hey how is it going | 19:48 |
-!- nube [~rho@49.244.67.162] has quit [Quit: Leaving.] | 19:50 | |
pickle27 | lisitsyn: good! | 19:59 |
pickle27 | did you see my question about the pos def matrix check? | 20:00 |
@lisitsyn | pickle27: yeah I don't think we have anything for that | 20:00 |
@lisitsyn | pickle27: do you know anything faster than computing eigenvalues btw? | 20:00 |
pickle27 | lisitsyn: I don't, I think that might be the only way? | 20:01 |
@lisitsyn | I guess so | 20:01 |
pickle27 | lisitsyn: what do you think about adding a member to SGMatrix that checks if it is pos def | 20:02 |
@lisitsyn | pickle27: I am ok with it | 20:03 |
pickle27 | lisitsyn: okay I might add one then | 20:03 |
pickle27 | it actually won't really help me where I needed it but it might come in handy later or for someone else | 20:03 |
pickle27 | are unit tests broken at the moment? | 20:06 |
@lisitsyn | pickle27: yeah a bit | 20:11 |
pickle27 | kk | 20:11 |
pickle27 | lisitsyn: sent a PR fixing Jade | 20:13 |
@lisitsyn | pickle27: I find JointDiago name a bit strange, do you? | 20:15 |
pickle27 | yeah I do | 20:15 |
@lisitsyn | pickle27: and this givens etc thing | 20:16 |
@lisitsyn | could you please correct the naming | 20:16 |
@lisitsyn | let me comment in code | 20:16 |
pickle27 | Im just not sure what else to call it | 20:16 |
pickle27 | its mostly a direct port from the authors release | 20:16 |
@lisitsyn | pickle27: may be just CJointDiagonalizer? | 20:17 |
pickle27 | perhaps | 20:18 |
pickle27 | we could also ask Andreas what he thinks we should call it | 20:18 |
@lisitsyn | sure why not | 20:18 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 20:37 | |
-!- iglesiasg [~Fernando@c83-251-227-64.bredband.comhem.se] has joined #shogun | 21:05 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 21:05 | |
@iglesiasg | sonney2k, lisitsyn, HeikoS : isn't it a big limitation that SGObjects can never be used in stack? | 21:14 |
@iglesiasg | wiking: ^ | 21:14 |
@lisitsyn | iglesiasg: like | 21:14 |
@lisitsyn | SGObject? | 21:14 |
@iglesiasg | lisitsyn: I am referring to https://github.com/shogun-toolbox/shogun/issues/1268 | 21:15 |
@lisitsyn | iglesiasg: well such thing disables polymorphism | 21:16 |
@iglesiasg | true | 21:17 |
@iglesiasg | lisitsyn: but one might one to have something in the stack for performance reasons | 21:17 |
@iglesiasg | lisitsyn: or is there no real difference in practice? | 21:18 |
@lisitsyn | no I don't think it is relevant | 21:18 |
@lisitsyn | you won't notice I believe | 21:18 |
@sonney2k | gsomix, you can use SG_REALLOC for that - maybe it is easiest to add a add_columns() / remove_columns() function to SGMatrix and then read in a few MB at a time | 21:33 |
@iglesiasg | gsomix: do you need to add and/or remove columns in a SGMatrix? Maybe the subset can help you to do that | 21:35 |
gsomix | iglesiasg, yeah, I need to add. for reading matrices from file. | 21:35 |
gsomix | good evening | 21:36 |
@iglesiasg | in my LMNN code I am using subsets to do something like this | 21:36 |
@iglesiasg | say you have a matrix like [1 2 3 4 5] | 21:36 |
@iglesiasg | where 1 2 3 ... denote full columns (with several rows), not just a number | 21:37 |
@iglesiasg | and you want to get stuff like | 21:37 |
@iglesiasg | [2 3 5] | 21:37 |
@sonney2k | iglesiasg, actually keeping things on stack is dangerous and calling for stack overflow errors | 21:38 |
@iglesiasg | sonney2k: really? isn't it like insane amount of variables in stack required for it to overflow? | 21:39 |
@lisitsyn | sonney2k: I can't imagine stack overflow on modern computers | 21:39 |
@sonney2k | iglesiasg, unfortunately not - I experienced that several times | 21:39 |
@iglesiasg | lisitsyn: with recursion is easy | 21:39 |
@lisitsyn | with reasonable code | 21:39 |
@sonney2k | no recursions | 21:39 |
@iglesiasg | sonney2k: I have experienced using recursion but not because of allocating variables | 21:39 |
@iglesiasg | and even with recursion it was because there was a bug :) | 21:40 |
@lisitsyn | well actually I'd love to see shogun objects as pimpl | 21:41 |
@lisitsyn | just wrappers for pointer to implementation | 21:41 |
@lisitsyn | like smartpointer but not a smartpointer | 21:41 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has joined #shogun | 21:43 | |
-!- lambday [67157e36@gateway/web/freenode/ip.103.21.126.54] has quit [Ping timeout: 250 seconds] | 22:10 | |
gsomix | sonney2k, ok, reading of vectors from csv works fine | 22:25 |
gsomix | today I have implemented StringReader - the class over SGVector<char> and CTokenizer with some useful methods like read_int(), etc. | 22:27 |
@sonney2k | gsomix, I had another idea - you could add a method to create a SGMatrix from a SGVector giving it some dimensions | 22:33 |
@sonney2k | gsomix, so you could just extend the vector all the time with resize | 22:33 |
gsomix | sonney2k, btw, number of rows I should define by first line in csv file, right? and number of lines then == number of columns | 22:38 |
gsomix | sonney2k, I like your idea. but now we don't have nor SGMatrix or SGVector in File interface - just pure arrays. | 22:41 |
@sonney2k | gsomix, there is no 'true' ordering for matrices. I guess | 22:45 |
@sonney2k | we need to support both | 22:45 |
@sonney2k | as in | 22:45 |
@sonney2k | 1 row == 1 column vector | 22:45 |
@sonney2k | and 1 row == 1 row vector | 22:45 |
gsomix | sonney2k, ok | 22:48 |
gsomix | need to sleep a little | 23:16 |
@sonney2k | metoo | 23:17 |
@sonney2k | cu | 23:17 |
@iglesiasg | good night people | 23:17 |
@iglesiasg | sonney2k: is it a nice Friday evening in your holidays place? | 23:17 |
@iglesiasg | I take that as a yes :D | 23:18 |
-!- van51 [~van51@athedsl-399972.home.otenet.gr] has quit [Quit: Leaving.] | 23:23 | |
-!- thoralf_ [~thoralf@37-5-32-132-dynip.superkabel.de] has joined #shogun | 23:58 | |
thoralf_ | Hey guys. | 23:58 |
-!- thoralf_ is now known as thoralf | 23:59 | |
@iglesiasg | Hi there! | 23:59 |
--- Log closed Sat Jul 20 00:00:35 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!