--- Log opened Tue Mar 04 00:00:06 2014 | ||
--- Day changed Tue Mar 04 2014 | ||
thoralf | Haha, cool. | 00:00 |
---|---|---|
dhruv13_ | i thought you were referring to that | 00:00 |
thoralf | I'm citing famous people. | 00:00 |
dhruv13_ | sadly, quite true | 00:00 |
dhruv13_ | i may doze off | 00:11 |
dhruv13_ | 219 mb done.... | 00:11 |
dhruv13_ | i'm napping | 00:11 |
@HeikoS | dhruv13_: haha :) | 00:12 |
@HeikoS | I am hacking Shogun ;) | 00:12 |
dhruv13_ | :D | 00:12 |
-!- radgaonkar [~rad@p4FC73B2E.dip0.t-ipconnect.de] has quit [Quit: leaving] | 00:14 | |
@HeikoS | dhruv13_: and? | 00:31 |
dhruv13_ | download done! | 00:37 |
dhruv13_ | HeikoS: yep, test fails | 00:43 |
@HeikoS | dhruv13_: nice | 00:45 |
@HeikoS | so thats good to validate and debug things then | 00:45 |
@HeikoS | dhruv13_: I will go home now, so see you tomorrow! good luck :) | 00:45 |
dhruv13_ | thanks!! | 00:45 |
dhruv13_ | i'll sleep(finally) | 00:46 |
dhruv13_ | bye!! | 00:46 |
@HeikoS | bye :) | 00:46 |
-!- dhruv13_ [~dhruv13@triband-mum-120.61.112.43.mtnl.net.in] has quit [Quit: Leaving] | 00:46 | |
lambday | HeikoS: still there?? | 00:47 |
@HeikoS | lambday: yeah | 00:47 |
@HeikoS | how are things? | 00:48 |
@HeikoS | lambday: still awake? :D | 00:48 |
@HeikoS | lambday: man its already late here ..... | 00:48 |
lambday | HeikoS: lol yeah! its 5:17 AM here :D | 00:48 |
@HeikoS | whoooooooooooo | 00:48 |
lambday | haha :D | 00:48 |
@HeikoS | lambday: so how are things? | 00:49 |
@HeikoS | hacking away all night? | 00:49 |
lambday | HeikoS: just to make sure, the way I changed was, to have two kernels for p and q which inits with features p and q (no appended features here) - for independence test | 00:49 |
lambday | HeikoS: and for getting kernel K, we add the subset using permutation | 00:49 |
lambday | HeikoS: while for L, we just call get_matrix - no subset needed | 00:50 |
lambday | since no permutation needed | 00:50 |
@HeikoS | yep! | 00:50 |
lambday | alright - then on the right track! | 00:51 |
@HeikoS | lambday: good stuff! | 00:51 |
lambday | HeikoS: sorry this PR is gonna be a huge one too :( name change makes it heavy :( | 00:51 |
lambday | HeikoS: btw we still would want to use precomputed kernel for L also, no? otherwise we gotta init it every time - | 00:52 |
@HeikoS | lambday: yeah definitely precomputed kernel | 00:53 |
@HeikoS | lambday: so there should be two modes | 00:53 |
@HeikoS | one is precompute all kernels | 00:53 |
@HeikoS | ah wait, is it actually possible to compute HSIC without precomputing kernel matrix? | 00:53 |
lambday | HeikoS: currently the code is there - only we're init-ing it each time | 00:54 |
lambday | these two modes are there already I mean | 00:54 |
@HeikoS | lambday: so if the kernel matrix is precomputed, one can just permute the entries | 00:55 |
lambday | HeikoS: yep! | 00:55 |
@HeikoS | I think in HSIC, kernel *is* precomputed currently | 00:55 |
@HeikoS | so no need to recompute any kernel elements, this is just done once | 00:55 |
lambday | HeikoS: and even though we don't need to permute the entries for the other matrix, its better we keep it precomputed, right? | 00:56 |
lambday | HeikoS: I mean, otherwise it anyway inits with the features | 00:56 |
@HeikoS | lambday: yeah | 00:56 |
@HeikoS | lambday: I mean for MMD it makes sense to not precompute the kernel to save memory | 00:57 |
@HeikoS | lambday: the quadratic time mmd is just a sum of kernel entries, so thats fine | 00:57 |
@HeikoS | but HSIC, let me check | 00:57 |
lambday | HeikoS: okay | 00:57 |
@HeikoS | lambday: so its the trace of some matrix product | 00:58 |
@HeikoS | so that requires to store the matrices | 00:58 |
@HeikoS | lambday: so precompute | 00:58 |
@HeikoS | lambday: in fact ,we can use Eigen3 here to compute the product in HSIC, thats easier to read :) | 00:58 |
lambday | HeikoS: yeah I was thinking about that | 00:59 |
lambday | HeikoS: although, not sure about the speed - cause we need to initialize all those eigen3 objects | 00:59 |
lambday | maps I meant | 00:59 |
@HeikoS | lambday: ah, we do the biased HSIC | 01:00 |
@HeikoS | or I did that in the code | 01:00 |
@HeikoS | lambday: so HSIC::compute_statistic can be done without storing the kernel matrix | 01:00 |
@HeikoS | I think in the MMD case, I did that two, but one can decide to use a CustomKernel with a preocmputed one, that will then avoid re-computing | 01:01 |
@HeikoS | and is important for the permutation test | 01:01 |
@HeikoS | lambday: yeah, checking Arthur code (see his website), he does that too | 01:03 |
@HeikoS | lambday: I will confirm with Dino tomorrow | 01:03 |
@HeikoS | lambday: but yeah, just compute the matrices for now | 01:03 |
@HeikoS | lambday: thats totally fine | 01:03 |
lambday | HeikoS: as in, keeping things as it was - except no subsetting for the second kernel | 01:04 |
@HeikoS | exactly | 01:04 |
lambday | HeikoS: alright - not sure whether its the current way - will send the PR by tomorrow - please have a look and comment then :) | 01:05 |
@HeikoS | lambday: will very happily do that, looking forward! | 01:06 |
lambday | HeikoS: I will read your thesis meanwhile to get basic ideas | 01:06 |
@HeikoS | lambday: yeah do that, its very simple though, not necessary to understand all the math, just basic ideas | 01:06 |
lambday | HeikoS: alright :) | 01:08 |
lambday | cool stuffs, man! | 01:08 |
@HeikoS | lambday: indeed! | 01:08 |
@HeikoS | too much good stuff | 01:08 |
@HeikoS | I am always too excited to sleep recently | 01:08 |
@HeikoS | currently getting a bit into STAN | 01:08 |
lambday | ah the mcmc project | 01:08 |
@HeikoS | I think we will use that to represent probability measures in Shogun | 01:08 |
@HeikoS | yeah | 01:09 |
@HeikoS | and then build our samplers on top, but we'll see | 01:09 |
@HeikoS | need to learn more first | 01:09 |
lambday | HeikoS: is roman applying this year? | 01:09 |
@HeikoS | lambday: unfortunately he is not a student anymore | 01:10 |
lambday | oh! :( | 01:10 |
@HeikoS | lambday: I am trying to push him to work on this anyways, but he has lots of things to do. | 01:10 |
@HeikoS | lambday: though he just speeded up the GPs by factor 2 :) | 01:10 |
lambday | great work!!! | 01:10 |
@HeikoS | lambday: ok I will go home now finally, | 01:12 |
@HeikoS | lambday: dont work too hard ;) | 01:12 |
@HeikoS | bye! | 01:12 |
lambday | HeikoS: lol you're still in lab :P | 01:12 |
@HeikoS | yeah | 01:12 |
lambday | HeikoS: yeah I will sleep on my keyboard now :d | 01:12 |
@HeikoS | haha | 01:12 |
lambday | HeikoS: ciao tomorrow :) | 01:12 |
lambday | tc | 01:12 |
@HeikoS | you too, bye! | 01:12 |
-!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 01:13 | |
lambday | HeikoS: bye :) | 01:13 |
-!- lambday [67157e4e@gateway/web/freenode/ip.103.21.126.78] has quit [Quit: Leaving.] | 01:51 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:05 | |
-!- bluenotebook [~bluenoteb@115.249.18.25] has joined #shogun | 02:30 | |
-!- zxtx [~zv@129-79-241-148.dhcp-bl.indiana.edu] has quit [Ping timeout: 240 seconds] | 02:56 | |
shogun-buildbot | build #633 of nightly_all is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_all/builds/633 | 03:22 |
-!- zxtx [~zv@c-98-223-196-32.hsd1.in.comcast.net] has joined #shogun | 03:37 | |
-!- soggyrusk [~bluenoteb@115.249.18.25] has joined #shogun | 04:22 | |
-!- soggyrusk [~bluenoteb@115.249.18.25] has left #shogun [] | 04:23 | |
-!- bluenotebook [~bluenoteb@115.249.18.25] has quit [Quit: kthxbye] | 04:23 | |
-!- claughinghouse1 [~claughing@pool-108-28-80-98.washdc.fios.verizon.net] has joined #shogun | 04:34 | |
-!- claughinghouse1 [~claughing@pool-108-28-80-98.washdc.fios.verizon.net] has quit [] | 04:57 | |
-!- lj [671b082d@gateway/web/freenode/ip.103.27.8.45] has joined #shogun | 06:39 | |
-!- lj [671b082d@gateway/web/freenode/ip.103.27.8.45] has left #shogun [] | 06:39 | |
-!- thoralf [~thoralf@91-65-138-84-dynip.superkabel.de] has quit [Quit: Konversation terminated!] | 07:59 | |
-!- sonne|osx [~sonne@89.204.154.71] has joined #shogun | 08:15 | |
-!- sonne|osx [~sonne@89.204.154.71] has quit [Ping timeout: 240 seconds] | 08:33 | |
-!- sonne|osx [~sonne@89.204.137.149] has joined #shogun | 08:35 | |
-!- sonne|osx [~sonne@89.204.137.149] has quit [Client Quit] | 08:35 | |
-!- lisitsyn [~lisitsyn@213.87.131.123] has quit [Read error: Connection reset by peer] | 09:13 | |
-!- huska1 [~huska@poutine.molgen.mpg.de] has quit [Read error: Connection reset by peer] | 09:59 | |
-!- lisitsyn [~lisitsyn@80.252.20.67] has joined #shogun | 10:03 | |
-!- sonne|work [~sonnenbu@24-134-74-216-dynip.superkabel.de] has left #shogun [] | 10:23 | |
-!- pirosb3 [8da395f8@gateway/web/freenode/ip.141.163.149.248] has joined #shogun | 10:30 | |
pirosb3 | Hello! goodmorning | 10:30 |
pirosb3 | could anyone help me with this error?? :) https://gist.github.com/PirosB3/9343151 | 10:31 |
@wiking | pirosb3: what's your arch+os? | 10:34 |
pirosb3 | OSX 10.7 64bit | 10:34 |
@wiking | mmm | 10:35 |
@wiking | interesting | 10:35 |
@wiking | i think you've just managed to step on a corner case again with the ever changing apis of osx | 10:35 |
@wiking | pirosb3: what's your clang version? | 10:35 |
pirosb3 | oh noooo! | 10:37 |
pirosb3 | Apple LLVM version 4.2 (clang-425.0.28) (based on LLVM 3.2svn) Target: x86_64-apple-darwin11.4.2 | 10:37 |
@wiking | mmm ok | 10:38 |
@wiking | pirosb3: fire an issue ... | 10:38 |
pirosb3 | of course. shall I put it on Github? | 10:38 |
pirosb3 | are we sure it's Apple's fault? and it's not mine?? | 10:38 |
@wiking | pirosb3: most probably it is | 10:39 |
pirosb3 | usually it's my fault :) | 10:39 |
pirosb3 | okay then, I will file an issue | 10:39 |
@wiking | pirosb3: but still put up an issue on github | 10:39 |
@wiking | at least it'll be logged and the next time something same happens then it's there + the solution | 10:39 |
@wiking | once we have the solution | 10:39 |
-!- dhruv13 [~dhruv13@triband-mum-120.61.60.174.mtnl.net.in] has joined #shogun | 10:47 | |
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has joined #shogun | 10:49 | |
@besser82 | sonney2k: just fired off some 3.2.0-build for fc19, fc20, rh ---> https://koji.fedoraproject.org/koji/packageinfo?packageID=17486 | 11:00 |
@besser82 | sonney2k: octave-3.8.0 problem is solved on rh now, thanks to orion :D | 11:01 |
pirosb3 | hey wiking: is this descriptive enought? anying I should add/change? https://github.com/shogun-toolbox/shogun/issues/1925 | 11:01 |
@besser82 | pirosb3: which SWIG-ver are you on? | 11:03 |
@wiking | pirosb3: almost... i've added a bit more debug info from the compilation error | 11:03 |
pirosb3 | swig: stable 2.0.12 | 11:03 |
@besser82 | pirosb3: just launched a build of 3.2.0-rel on Fedora with SWIG 2.0.12-rel; let's see if it happenes here, too... ;) | 11:04 |
@wiking | besser82: it's osx related... namely most probably the dirent hacks that we had to do because of osx's constant change of dirent api | 11:05 |
@wiking | besser82: so building with fedora will not cause anything similar | 11:05 |
pirosb3 | besser82: are you sure it's library related | 11:05 |
@wiking | pirosb3: what's the exact version of your osx btw | 11:06 |
@wiking | apart from 10.7 | 11:06 |
pirosb3 | it looks like incompatibility of interfaces | 11:06 |
pirosb3 | wiking: sure, ill update the post | 11:06 |
@wiking | 10.7.x? | 11:06 |
@besser82 | wiking: ah, ok, thx! didn't know that... You know I don't use stuff with bitten fruits on it's case. :D | 11:06 |
pirosb3 | 7.5 | 11:06 |
@wiking | pirosb3: 10.7.5 ? | 11:07 |
pirosb3 | Yes wiking updated the issue | 11:07 |
pirosb3 | any suggestings I could try out now wiking? | 11:08 |
-!- lambday [67157f4e@gateway/web/freenode/ip.103.21.127.78] has joined #shogun | 11:08 | |
@wiking | mmm well there's a big macro if/else in src/shogun/io/SGIO.h | 11:08 |
@wiking | that actually sets the type of dirent struct | 11:08 |
@wiking | startinf from line 74 | 11:08 |
pirosb3 | wiking: :oh yes I can see it | 11:09 |
@wiking | so my guess is that maybe you would need to change there something | 11:09 |
@wiking | that actually will set your dirent struct | 11:09 |
@wiking | to the right value | 11:09 |
pirosb3 | yes | 11:09 |
@wiking | actually i dont even get | 11:10 |
@wiking | why you have | 11:10 |
@wiking | const dirent * | 11:10 |
@wiking | when you should have | 11:10 |
pirosb3 | yes, I changed it. building :) | 11:10 |
pirosb3 | oh no! it will take ages again :) | 11:11 |
@wiking | so basically put it in a way that your's is const dirent as well | 11:11 |
@wiking | like it's in 10.8 or later | 11:11 |
pirosb3 | once I solve this, shall I update the issue? | 11:12 |
@wiking | well yeah at least put it there if this actually solved your problem or not | 11:12 |
pirosb3 | ok. Once this is solved, is it worth thinking of a persistent solution and making a PR back? | 11:13 |
@wiking | of course | 11:14 |
pirosb3 | hey guys, does anyone know AdaBoost? | 11:27 |
pirosb3 | when would I use AdaBoost, compared to other classifiers? | 11:28 |
-!- rad_ [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has joined #shogun | 11:41 | |
-!- rad_ [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has quit [Client Quit] | 11:41 | |
pirosb3 | hey wiking, has anyone started working on the SoC deep learning task? | 11:51 |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 11:52 | |
@wiking | pirosb3: i think there were already some people | 11:54 |
@wiking | pirosb3: check the mailing list archive | 11:54 |
pirosb3 | ok. so you only get 1 person per task right? | 11:54 |
-!- radgaonkar [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has joined #shogun | 12:14 | |
-!- pirosb3 [8da395f8@gateway/web/freenode/ip.141.163.149.248] has quit [Quit: Page closed] | 12:24 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has joined #shogun | 12:25 | |
-!- cameo54321 [cb523105@gateway/web/freenode/ip.203.82.49.5] has joined #shogun | 12:42 | |
@besser82 | wiking: any know reason for deb3-buildbot stopped around Feb. 17th ??? | 12:44 |
@besser82 | sonney2k: ^^^^ | 12:44 |
-!- radgaonkar [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has quit [Ping timeout: 244 seconds] | 12:47 | |
dhruv13 | sonney2k: hey! I'm having problems running the integration tests after a fresh build-install; | 13:13 |
-!- sunil1337 [0e8bd055@gateway/web/freenode/ip.14.139.208.85] has joined #shogun | 13:13 | |
dhruv13 | sonney2k: i get the following error Traceback (most recent call last): | 13:14 |
dhruv13 | File "./tester.py", line 3, in <module> | 13:14 |
dhruv13 | import modshogun | 13:14 |
dhruv13 | File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 30, in <module> | 13:14 |
dhruv13 | _modshogun = swig_import_helper() | 13:14 |
dhruv13 | File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 26, in swig_import_helper | 13:14 |
dhruv13 | _mod = imp.load_module('_modshogun', fp, pathname, description) | 13:14 |
dhruv13 | ImportError: /usr/local/lib/python2.7/dist-packages/_modshogun.so: undefined symbol: _ZN6shogun9CSGObject6equalsEPS0_d | 13:14 |
PirosB3 | hi @dhruv13 | 13:15 |
dhruv13 | besser82: tester.py fails altogether after rebuild... | 13:15 |
dhruv13 | PirosB3: hi! | 13:15 |
sunil1337 | hi guys.! | 13:16 |
PirosB3 | hi @sunil1337 | 13:16 |
sunil1337 | the build also fails without eigen3 lib | 13:16 |
dhruv13 | sunil1337: hello! | 13:16 |
sunil1337 | dhruv13: hey bro | 13:16 |
PirosB3 | @sunil1337 did you specify install of eigen3 with cmake? | 13:16 |
sunil1337 | PirosB3: no.. its not #defined | 13:16 |
sunil1337 | i've found the cause | 13:17 |
PirosB3 | oh ok | 13:17 |
sunil1337 | its ~project-dir/src/shogun/base/class_list.cpp fails to build. because the class CPCA is not defined when there is no eigen3 lib | 13:18 |
sunil1337 | but it tries to return a new object of CPCA type at line 604 | 13:19 |
PirosB3 | @sunil1337 I think I had that error too | 13:19 |
PirosB3 | @sunil1337 can you send me the cmake command you ran? | 13:19 |
PirosB3 | @besser82 is there any doc that explains the shogun "pipeline"? | 13:20 |
sunil1337 | PirosB3: ccmake -DPythonModular=ON rest it tries to find eigen3 when it doesn't find it it doesn't #define HAVE_EIGEN3 | 13:21 |
PirosB3 | ok cool | 13:22 |
@besser82 | PirosB3: which "Pipeline"? | 13:22 |
dhruv13 | PirosB3, sunil1337: why don't you install Eigen3 and re-run cmake? that should fix the build | 13:22 |
sunil1337 | dhruv13: thats what i did :D | 13:23 |
dhruv13 | didn't work? | 13:23 |
PirosB3 | @sunil1337 cmake -DBUNDLE_EIGEN=ON -DBUNDLE_JSON=ON | 13:23 |
PirosB3 | you need to enable EIGEN | 13:24 |
dhruv13 | did you specify a different install directory than /usr/local? | 13:24 |
PirosB3 | try adding these params | 13:24 |
sunil1337 | dhruv13: it worked..but the thing is eigen3 is optional | 13:24 |
dhruv13 | sunil1337: yeah... that's a little weird | 13:24 |
PirosB3 | @dhruv13 I think its necessary for setting up developer | 13:25 |
dhruv13 | yeah | 13:25 |
PirosB3 | @besser82 shogun has many kernels, and it also has many classification/regression techniques | 13:26 |
PirosB3 | I found a list of all of them, but I am trying to think of a way to combine them in an API | 13:26 |
sunil1337 | PirosB3: can u please pass it on | 13:32 |
PirosB3 | @sunil1337 regarding what? | 13:33 |
sunil1337 | PirosB3: the list | 13:35 |
PirosB3 | oh sure :) as soon as I have it :) | 13:36 |
sunil1337 | PirosB3: :D sure | 13:36 |
sunil1337 | gotta go ... bye guys. | 13:36 |
-!- sunil1337 [0e8bd055@gateway/web/freenode/ip.14.139.208.85] has quit [Quit: Page closed] | 13:36 | |
PirosB3 | sure | 13:36 |
PirosB3 | see ya sunhill | 13:36 |
PirosB3 | sunil | 13:37 |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 13:37 | |
PirosB3 | @besser82 are there any Go bindings for Shogun? | 13:45 |
@besser82 | PirosB3: sry, been afk for lunch :( | 13:45 |
-!- lisitsyn1 [~lisitsin@mxs.kg.ru] has quit [Read error: Connection reset by peer] | 13:45 | |
PirosB3 | ahah no worries | 13:45 |
-!- lisitsyn2 [~lisitsin@mxs.kg.ru] has joined #shogun | 13:45 | |
@besser82 | PirosB3: there are currently no "Go" bindings for SHOGUN avail... | 13:45 |
PirosB3 | ok | 13:45 |
@besser82 | PirosB3: How about using Ruby? | 13:45 |
PirosB3 | @besser82 both Ruby and Python have the binding. I am just concerned about concurrency level | 13:46 |
PirosB3 | Golang is fast and is really good for concurrency. Is it easy to serialize classifiers in shogun? | 13:47 |
@besser82 | PirosB3: they should be easily serializable like with hdf5 or json, which is both supported by shogun directly | 13:47 |
PirosB3 | I would discard JSON, something binary would be better | 13:48 |
PirosB3 | it needs to be fast. As it is an API, people could be using multiple classifiers | 13:48 |
@besser82 | PirosB3: then go for hdf5 ;) | 13:49 |
PirosB3 | In my head, a classifier is spun up and kept in memory for a specific amount of time, then is deallocated | 13:49 |
@besser82 | PirosB3: you should be able to serialize any obj deriving from SGObject from / to hdf5 ;) | 13:49 |
PirosB3 | ok, can that be also done with the Python interface?= | 13:50 |
@besser82 | PirosB3: afaik, yes; with all *_modular | 13:51 |
PirosB3 | I have not still had the chance to install shogun! it's always failing :( even if I change the definition of CONST_DIRENT_T. It blows up later on | 13:51 |
@besser82 | PirosB3: with which error? | 13:51 |
PirosB3 | a c directory api call | 13:52 |
PirosB3 | I can't remember what it was :( | 13:52 |
PirosB3 | I am really considering to recompile everyone in VM | 13:53 |
PirosB3 | ah yeah! scandir | 13:53 |
@besser82 | PirosB3: could please try to reproduce and give me some more verbose error-msg? ;) | 13:57 |
PirosB3 | sure | 13:58 |
PirosB3 | I'll be back soon guys, thanks for your help | 14:00 |
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 14:00 | |
dhruv13 | sonney2k: swig woes :( | 14:08 |
dhruv13 | Traceback (most recent call last): | 14:08 |
dhruv13 | File "./tester.py", line 3, in <module> | 14:08 |
dhruv13 | import modshogun | 14:08 |
dhruv13 | File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 30, in <module> | 14:08 |
dhruv13 | _modshogun = swig_import_helper() | 14:08 |
dhruv13 | File "/usr/local/lib/python2.7/dist-packages/modshogun.py", line 26, in swig_import_helper | 14:08 |
dhruv13 | _mod = imp.load_module('_modshogun', fp, pathname, description) | 14:08 |
dhruv13 | ImportError: /usr/local/lib/python2.7/dist-packages/_modshogun.so: undefined symbol: _ZN6shogun9CSGObject6equalsEPS0_d | 14:08 |
-!- radgaonkar [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has joined #shogun | 14:24 | |
-!- cameo54321 [cb523105@gateway/web/freenode/ip.203.82.49.5] has quit [Ping timeout: 245 seconds] | 14:44 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 15:42 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Client Quit] | 15:43 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has joined #shogun | 15:47 | |
-!- lisitsyn2 [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 16:28 | |
-!- dhruv13 [~dhruv13@triband-mum-120.61.60.174.mtnl.net.in] has quit [Ping timeout: 264 seconds] | 16:51 | |
@besser82 | sonney2k: currently building recent shogun git-snap on rawhide :D https://koji.fedoraproject.org/koji/taskinfo?taskID=6595206 | 16:52 |
-!- dhruv13 [~dhruv13@triband-mum-120.61.127.208.mtnl.net.in] has joined #shogun | 17:04 | |
@besser82 | anyone else ever hit a happening like this? ---> http://abstrusegoose.com/34 | 17:17 |
-!- siddharth_sharma [0e8bf3a2@gateway/web/freenode/ip.14.139.243.162] has joined #shogun | 17:19 | |
-!- sm004 [67157e4c@gateway/web/freenode/ip.103.21.126.76] has joined #shogun | 17:28 | |
siddharth_sharma | hello, i am trying to build shogun but when i am running ./configure --interface= in /shogun/src dir it shows ./configure no such file or directory exists | 17:31 |
siddharth_sharma | i have made sure i have all utilities t | 17:33 |
sm004 | Hi, newbie here. How do I run the unit tests? | 17:33 |
-!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:37 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 17:37 | |
PirosB3 | Hey guys, can anyone give me a heads up on my SoC idea before I post it?? :) thanks | 17:41 |
PirosB3 | https://gist.github.com/PirosB3/9350092 | 17:41 |
@HeikoS | PirosB3: hi! | 17:41 |
PirosB3 | @HeikoS Hi! how are you? | 17:41 |
@HeikoS | PirosB3: good thanks! how are you? | 17:42 |
@HeikoS | PirosB3: currently looking at your text, what is the rest API? | 17:42 |
PirosB3 | @HeikoS the rest API would be a way to interact with Shogun as a SaaS | 17:43 |
@HeikoS | PirosB3: could you elaborate a bit on that? I dont know it ;) | 17:43 |
PirosB3 | @HeikoS an example would be to train a classifier: you would do a call to POST /classifiers/train { training_label: "/sentiments/train.svmlight", testing_label: "sentiment/test.svmlight, kernel: "gaussian", ... } | 17:45 |
siddharth_sharma | hello, i am trying to build shogun but when i am running ./configure --interface= in /shogun/src dir it shows ./configure no such file or directory exists | 17:45 |
@HeikoS | PirosB3: do you got any reference on this? that I could read up about? | 17:45 |
PirosB3 | @HeikoS thanks for the comments, I will elaborate on it and update | 17:45 |
@HeikoS | siddharth_sharma: pls build Shogun using cmake | 17:45 |
@HeikoS | PirosB3: I dont really give any comments yet right? ;) | 17:45 |
@HeikoS | PirosB3: but the docker stuff (an automated image) would be appreciated | 17:45 |
PirosB3 | @HeikoS look at this | 17:46 |
PirosB3 | https://developers.google.com/prediction/docs/reference/v1.6/#Trainedmodels | 17:46 |
siddharth_sharma | @HeikoS: ok | 17:46 |
PirosB3 | @HeikoS I want to reshape everything to make sure it is as understandable as possible, then I will post it | 17:46 |
@HeikoS | PirosB3: I am not too sure whether thats something feasible to do for us | 17:46 |
PirosB3 | @HeikoS what are the concerns? | 17:47 |
@HeikoS | PirosB3: we already have problems maintaining existing stuff, adding another interface makes life even more complicate | 17:47 |
@HeikoS | PirosB3: did you already talk to someone about this?, wiking? | 17:47 |
PirosB3 | I have already spoken to wiking | 17:47 |
PirosB3 | He told me to post it, and see the reaction. He looked happy about it | 17:47 |
@HeikoS | PirosB3: what is the state of the discussion? | 17:47 |
sm004 | Could someone help me with running the unit tests? | 17:47 |
@wiking | HeikoS: yo | 17:48 |
@HeikoS | wiking: tell me about this SaaS stuff | 17:48 |
PirosB3 | See, this is a very early idea. I want to make sure it is feasable | 17:48 |
@HeikoS | sm004: whats the problem | 17:48 |
PirosB3 | wiking: I want to post this https://gist.github.com/PirosB3/9350092 | 17:48 |
@wiking | PirosB3: just send it to the mailing list | 17:49 |
PirosB3 | wiking: is it clear enough? | 17:49 |
@wiking | PirosB3: suuuuure! | 17:49 |
PirosB3 | cool thanks guys. Sending now | 17:49 |
@wiking | PirosB3: and anyway if it's not clear enough people will start asking questions | 17:49 |
@wiking | so it'll get iterated couple of times for sure | 17:49 |
@wiking | HeikoS: so what do u think of the exchangeable eigen solver? | 17:49 |
@HeikoS | wiking: how is the notebook bot coming along? would be great to have | 17:50 |
@HeikoS | wiking: nice idea! | 17:50 |
@wiking | HeikoS: well that'd be an extension of shogun cloud | 17:50 |
@HeikoS | wiking: but I was hoping that eigen3 can be replaced with something like that | 17:50 |
@HeikoS | wiking: besser82 once showed this lib which has eigen3 api but dies stuff on the GPU /in parallel | 17:50 |
@wiking | HeikoS: that it not only allows u to run your own ipython notebook server but you can have like a REST api where u can do training, classifying etc via simple rest api calls | 17:50 |
-!- sm004_ [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 17:50 | |
@HeikoS | wiking: without changing calls | 17:50 |
@wiking | HeikoS: viennacl ;) | 17:51 |
@besser82 | wiking, | 17:51 |
@wiking | HeikoS: that actually does those things smoothly with eigen based matrices etc | 17:51 |
@wiking | besser82: yes? | 17:51 |
@HeikoS | wiking: I like the idea of this REST a lot, but I am unsure about feasability here, thats complicated stuff, and currently more urgent things are not working/existing: web-demos, parallel framework, static interfaces, etc | 17:51 |
@besser82 | wiking, HeikoS: viennacl can replace lapack / blas too ;) | 17:52 |
-!- sm004 [67157e4c@gateway/web/freenode/ip.103.21.126.76] has quit [Ping timeout: 245 seconds] | 17:52 | |
@HeikoS | wiking: ok, so why do we need to give a seperate solver then? | 17:52 |
@besser82 | wiking, HeikoS: possibly atlas as well ;) | 17:52 |
@HeikoS | wiking: cant we just replace the backend globally? | 17:52 |
@wiking | HeikoS: well just because in PCA atm the solver is hardcoded | 17:52 |
@wiking | HeikoS: so it'd be cool to have something like PCA::set_eigen_solver(EigenSolver*) | 17:52 |
@wiking | HeikoS: and then that solver is being used in PCA for solving the eigen decomposition | 17:53 |
@HeikoS | wiking: I dont get it, I mean we currently do an eigen3 call compute eigenvalues | 17:53 |
@HeikoS | wiking: yeah I agree, but we do that right? its an eigen call | 17:53 |
@HeikoS | SelfAdjointEigenSolver<MatrixXd> eigenSolve = | 17:53 |
@HeikoS | SelfAdjointEigenSolver<MatrixXd>(cov_mat); | 17:53 |
@HeikoS | eigenValues = eigenSolve.eigenvalues().tail(max_dim_allowed); | 17:53 |
@wiking | HeikoS: yes...but it's specifially calling one given eigen solver | 17:53 |
@HeikoS | wiking: ah I see | 17:53 |
@HeikoS | wiking: so I thought changin to vienna stuff automatically changes that | 17:54 |
@wiking | i'd say (i dont know for 100%) that there's different eigen solvers in eigen3 as well | 17:54 |
@wiking | HeikoS: noup | 17:54 |
@wiking | HeikoS: that's a different solver... | 17:54 |
@wiking | that's why i want to have the eigen solver in PCA a parameter | 17:54 |
@HeikoS | wiking: I see | 17:54 |
@HeikoS | wiking: what about an enum? | 17:55 |
@wiking | mmm | 17:55 |
@HeikoS | wiking: but in fact you are right, this should be shared along all shogun | 17:55 |
@HeikoS | so all eigenproblems should use an eigensolver | 17:55 |
@wiking | brb (30seconds) | 17:55 |
@HeikoS | ok | 17:55 |
PirosB3 | @wiking @besser82 did the mail come through? Im not sure it was posted correctly :) | 17:55 |
@HeikoS | wiking: there is some sparse eigensolver base classes already | 17:55 |
@besser82 | PirosB3: yes, I got it :D | 17:55 |
PirosB3 | cool! | 17:55 |
@HeikoS | lambday: around? | 17:58 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 17:59 | |
shogun-notifier- | shogun: lambday :develop * f9a815d / / (41 files): https://github.com/shogun-toolbox/shogun/commit/f9a815dd38652d254c905f68d8e8cb7d65c85c4c | 17:59 |
shogun-notifier- | shogun: hsic bugfix and interface changes | 17:59 |
shogun-notifier- | shogun: Heiko Strathmann :develop * e04a5bd / / (41 files): https://github.com/shogun-toolbox/shogun/commit/e04a5bdd5d87463886e157af8792a70508a95c80 | 17:59 |
shogun-notifier- | shogun: Merge pull request #1926 from lambday/feature/hsic_bugfix | 17:59 |
shogun-notifier- | shogun: | 17:59 |
shogun-notifier- | shogun: hsic bugfix and interface changes | 17:59 |
-!- dhruv13 [~dhruv13@triband-mum-120.61.127.208.mtnl.net.in] has quit [Quit: Leaving] | 18:00 | |
@wiking | HeikoS: and yeah actually this should be covered by /mathematics/linalg/eigsolver | 18:02 |
@HeikoS | wiking: so we want to have a shared eigensolver base class which implements different backends for that, good idea! | 18:02 |
@wiking | HeikoS: becauase i suppose other parts (not only PCA) would benefit of a GPU based eigen solver... | 18:02 |
@HeikoS | wiking: yeah, the current eigensolver base class can even be used (with a few minor simplifications) | 18:02 |
-!- lambday [67157f4e@gateway/web/freenode/ip.103.21.127.78] has quit [Ping timeout: 245 seconds] | 18:02 | |
@wiking | HeikoS: yeah that class should be refactored a bit | 18:03 |
@HeikoS | wiking: the current one is for sparse eigenvalue problems where one wants min and max only | 18:03 |
@HeikoS | wiking: do you want to write an issue on that? | 18:03 |
@wiking | HeikoS: but that would be a good base to have various eigen solvers | 18:03 |
@HeikoS | wiking: yeah I agree | 18:03 |
@wiking | HeikoS: yeah i'll put up an entrance task for that | 18:03 |
@HeikoS | wiking: they should all work on linear operators, in this case dense linear operator | 18:03 |
shogun-buildbot | build #2725 of deb1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2725 | 18:03 |
@wiking | HeikoS: yep.... | 18:03 |
@HeikoS | wiking: we have classes for that already, but there should be convinience methods to directly solve SGMatrix Eigenproblems | 18:03 |
@HeikoS | and then reference all Shogun eigenproblems to that | 18:03 |
@HeikoS | wiking: nice! | 18:04 |
@HeikoS | wiking: what about making this class global, i.e. one can change the dense eigensolver globally | 18:04 |
@HeikoS | with one instruction | 18:04 |
@wiking | HeikoS: what do u mean globally? | 18:04 |
@HeikoS | or is that too unflexible? | 18:04 |
@HeikoS | wiking: like two classes use a dense eigensolver | 18:04 |
@wiking | and? | 18:05 |
@HeikoS | wiking: and to change both of them from eigen3 to vienna, I just need to set one flag rather than changing in both implementations seperately | 18:05 |
@wiking | ah | 18:05 |
@wiking | mmm | 18:05 |
@wiking | well i'll write the task liek that | 18:05 |
@wiking | ;) | 18:05 |
@wiking | so for example | 18:05 |
@wiking | if viennacl is available | 18:05 |
@wiking | then it'd use viennacl based solvers by default | 18:05 |
@HeikoS | wiking: yeah or even automatically | 18:05 |
@wiking | yeah yeah | 18:05 |
@HeikoS | cool | 18:05 |
@HeikoS | love that!° | 18:06 |
@HeikoS | wiking: do you know stan? | 18:06 |
@wiking | k i'll post the issue | 18:06 |
@wiking | HeikoS: no i dont think so | 18:06 |
@wiking | HeikoS: btw: bot i'll work on it now... sorry too many things at once everywhere... | 18:06 |
@HeikoS | wiking: I would like to depend some things on that, but it depends on eigen3, and boost, I wonder whether thats too much, its also not in debian or so | 18:06 |
@HeikoS | wiking: cool! appreciated | 18:06 |
@wiking | HeikoS: gimme a link | 18:06 |
@wiking | HeikoS: heheh i remember the time when sonney2k was very against boost... | 18:07 |
@wiking | boost is heavy weight | 18:07 |
@HeikoS | wiking: well | 18:07 |
@HeikoS | https://github.com/stan-dev/stan | 18:07 |
-!- sm004_ [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [Ping timeout: 245 seconds] | 18:07 | |
@HeikoS | wiking: I refuse to implement basic things, thats stupid | 18:07 |
@HeikoS | I dont like the idea of integrating code into Shogun, rather link | 18:07 |
@HeikoS | wiking: and I dont like to re-implement cdf of distributions | 18:08 |
@HeikoS | wiking: look up tables and other horrible stuff | 18:08 |
@HeikoS | wiking: so if user wants to use stuff, should install proper tools | 18:08 |
@HeikoS | wiking: but yeah whatever | 18:08 |
@wiking | HeikoS: well talk about this with sonney2k | 18:08 |
@HeikoS | wiking: will do, its not priority currently | 18:08 |
@HeikoS | wiking: notebook more important :) | 18:08 |
@wiking | ehhehe ok i got it | 18:08 |
@HeikoS | wiking: current git passes travis but fails notebooks | 18:08 |
@wiking | HeikoS: ok got it | 18:09 |
@HeikoS | wiking: nice, thanks a lot! :) | 18:09 |
@wiking | nw | 18:09 |
@wiking | bbl | 18:09 |
-!- radgaonkar [~rad@desktop-242.kma.informatik.tu-darmstadt.de] has quit [Quit: leaving] | 18:10 | |
@HeikoS | wiking: just out of curiosity how does thi coverage stuff work? | 18:17 |
-!- zxtx [~zv@c-98-223-196-32.hsd1.in.comcast.net] has quit [Ping timeout: 264 seconds] | 18:18 | |
@HeikoS | wiking: in fact, we should do the same thing for SVD as for EVD | 18:18 |
@HeikoS | wiking: and for linear solvers | 18:18 |
@HeikoS | wiking: and matrix multiplications | 18:18 |
@HeikoS | wiking: at least matrix-matrix | 18:19 |
@wiking | HeikoS: yep... indeed | 18:21 |
@wiking | HeikoS: coverage is based on examples and unit tests | 18:21 |
@wiking | HeikoS: basically it shows the percentage of code in shogun being actually used/called by unit tests and examples | 18:21 |
@HeikoS | wiking: yeah, but how is done evaluated? | 18:22 |
@wiking | HeikoS: so atm 28% of shogun library is not being called at all | 18:22 |
@HeikoS | wiking: cool btw! | 18:22 |
@HeikoS | wiking: hehe, well one doesnt need 100% | 18:22 |
@HeikoS | wiking: more important to think of meaningful tests | 18:22 |
shogun-notifier- | shogun: Roman Votyakov :develop * 0a9bba3 / / (14 files): https://github.com/shogun-toolbox/shogun/commit/0a9bba336487f6b06c8df119aa7bfb0def98019b | 18:23 |
shogun-notifier- | shogun: fix parameters hashing framework | 18:23 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 8893a8f / / (14 files): https://github.com/shogun-toolbox/shogun/commit/8893a8f0537e44dd89270ff7c45f208bc4566355 | 18:23 |
shogun-notifier- | shogun: Merge pull request #1922 from votjakovr/feature/fix_parameter_hashing | 18:23 |
shogun-notifier- | shogun: | 18:23 |
shogun-notifier- | shogun: Fix parameter hashing framework | 18:23 |
@wiking | HeikoS: yes and no.... i mean it would be greate to get around 90% coverage at least... but of course it's hard with all the #ifdefs ... | 18:23 |
@wiking | HeikoS: since basically now what is saying is if we take the result as is that 28% of the code in shogun is not ever being tested or used | 18:24 |
@HeikoS | wiking: yeah | 18:24 |
@wiking | and that's bad | 18:24 |
@HeikoS | wiking: remember 1 year ago? :) | 18:24 |
@HeikoS | 0% | 18:24 |
@HeikoS | ah no we at least had examples :) | 18:24 |
@wiking | well i would say 50% or something | 18:24 |
@wiking | since we had examples... | 18:24 |
@HeikoS | wiking: important are the unit test of low level modules | 18:24 |
@HeikoS | failing examples are quite meaningless | 18:24 |
@HeikoS | and no check for wrong results | 18:24 |
@wiking | yeah those were necessary.... | 18:25 |
@HeikoS | I love the unit tests | 18:25 |
-!- lambday [67157f4e@gateway/web/freenode/ip.103.21.127.78] has joined #shogun | 18:25 | |
lambday | HeikoS: there?? | 18:25 |
lambday | HeikoS: went for dinner.. just came back | 18:26 |
lambday | HeikoS: checking your comments | 18:26 |
@HeikoS | lambday: hey! | 18:26 |
@HeikoS | lambday: its merged, currently reading docs | 18:26 |
lambday | HeikoS: alright - please let me know if anything is inconsistent | 18:27 |
lambday | HeikoS: I will surely make the notebook work in the next PR... this was getting way too big - so didn't include that in this one | 18:27 |
@HeikoS | lambday: cool! we will soon have some travis like checks on the notebooks | 18:28 |
@HeikoS | lambday: yeah its big stuff, but cannot avoid that | 18:28 |
lambday | HeikoS: ah that would be great! | 18:28 |
lambday | HeikoS: yeah I mentioned that in my comment :P | 18:28 |
-!- siddharth_sharma [0e8bf3a2@gateway/web/freenode/ip.14.139.243.162] has quit [Ping timeout: 245 seconds] | 18:30 | |
@HeikoS | lambday: I think the subclasses should not repeat the class comments from the base class | 18:34 |
@HeikoS | lambday: just say "same as blabla" and state the additions | 18:34 |
@HeikoS | lambday: mmh | 18:35 |
@HeikoS | lambday: or just leave it, its nice in fact :) | 18:35 |
lambday | HeikoS: its a bit of repeating but doesn't harm I suppose? | 18:35 |
@HeikoS | lambday: yeah true! :) | 18:36 |
lambday | I am also compiling the doc | 18:36 |
lambday | :D | 18:36 |
@HeikoS | lambday: just read all of them, all nice! | 18:37 |
@HeikoS | very nice and clear now, check the class diagram when you open CHypothesisTest | 18:37 |
lambday | HeikoS: checking | 18:37 |
lambday | :D | 18:37 |
lambday | argh this git flow feature finish!!! | 18:37 |
@HeikoS | lambday: btw google for last years NIPS B-Test paper we might want to add that too in the project | 18:38 |
lambday | HeikoS: B-test: A Non-parametric, Low Variance Kernel Two-sample Test - this one? | 18:39 |
@HeikoS | lambday: yep! | 18:39 |
@HeikoS | kinda cool | 18:39 |
lambday | HeikoS: awesome - checking | 18:39 |
@HeikoS | tradeoff between quadratic time and linear time mmd that combined advantages ofd both | 18:39 |
@HeikoS | lambday: but not priority | 18:39 |
@HeikoS | lambday: but ill put up an issue anyways ;) | 18:40 |
lambday | HeikoS: okay - time to read it up! :) | 18:40 |
lambday | HeikoS: I thought of going through your report in next day or two | 18:41 |
lambday | HeikoS: other stuffs are sort of research based, right? unequal number of samples part, etc.. | 18:42 |
@HeikoS | lambday: ah the kernel selection ones? | 18:42 |
@HeikoS | yeah thats nice | 18:42 |
@HeikoS | lambday: unqeual number of samples is easy for quadratic time MMD | 18:42 |
@HeikoS | for linear, its not really needed | 18:42 |
lambday | HeikoS: okay - let me read up the basics - now that I have messed up a bit with the actual code, it would help a bit I suppose | 18:43 |
lambday | HeikoS: btw are you happy the way permutation is handled with the new implementation? | 18:44 |
@HeikoS | lambday: there are the graphical examples | 18:44 |
@HeikoS | lambday: and the notebook is very incomplete, feel free to extend that, add better explanaitions and case studies | 18:44 |
lambday | HeikoS: alright - next task, as I go through with the concepts, I will try to enrich the notebook | 18:45 |
lambday | HeikoS: the graphical examples are already there - I thought of including something new - may be incorporate google translator for the different language but same context example | 18:46 |
lambday | haven't though much about it, but let me think | 18:46 |
@HeikoS | lambday: these examples should be moved into the notebook and then be deleted | 18:47 |
lambday | HeikoS: alrighty | 18:47 |
@HeikoS | lambday: https://github.com/shogun-toolbox/shogun/issues/1928 | 18:49 |
@HeikoS | lambday: large one :) | 18:49 |
@HeikoS | more part of the projects | 18:49 |
lambday | HeikoS: wow!... | 18:49 |
lambday | HeikoS: better getting my head into reading the paper | 18:50 |
@HeikoS | lambday: its very simple | 18:51 |
@HeikoS | lambday: just compute the quadratic MMD on diagonal blocks of the kernel matrix, and then average over them | 18:51 |
@HeikoS | ah let me correct something | 18:51 |
@HeikoS | lambday: ok | 18:55 |
@HeikoS | ok gotta run off now, be back later, | 18:55 |
@HeikoS | see you | 18:55 |
lambday | HeikoS: alright - having a look at the paper | 18:55 |
lambday | HeikoS: ciao | 18:55 |
-!- HeikoS [~heiko@pat-191-250.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 18:57 | |
-!- HeikoS [~heiko@nat-169-252.internal.eduroam.ucl.ac.uk] has joined #shogun | 18:57 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 18:57 | |
-!- HeikoS [~heiko@nat-169-252.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 269 seconds] | 19:03 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 19:03 | |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 19:07 | |
-!- zxtx [~zv@ip-64-134-197-166.public.wayport.net] has joined #shogun | 19:09 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 19:24 | |
-!- dhruv13 [~dhruv13@84.64.13.210] has joined #shogun | 19:54 | |
@sonney2k | dhruv13, I read the link you send me again about the nan equals - and java's .equals() indeed returns true for nan but not for = - so IMHO we should return true for nan on equals too | 19:56 |
dhruv13 | sonney2k: there is an isNAN which does that, but I get your point... I'll handle the case specially | 19:57 |
dhruv13 | sonney2k: once i figure out why the integration test fails... | 19:57 |
dhruv13 | sonney2k: if i make changes to the CSGObject::equals() function prototype, the python bindings should get updated automatically right? | 20:00 |
-!- thoralf [~thoralf@91-66-35-191-dynip.superkabel.de] has joined #shogun | 20:00 | |
thoralf | Hey GUIs. | 20:00 |
dhruv13 | hello! | 20:01 |
thoralf | Hey dhruv13, how is it going? | 20:02 |
dhruv13 | not too bad i guess... a little stumped | 20:02 |
thoralf | dhruv13: Sounds you're stuck. Need help? | 20:02 |
dhruv13 | in a while ;-) | 20:03 |
dhruv13 | i'm actually travelling, so might go off any time | 20:03 |
thoralf | If you're stuck, don't struggle too long. Just try to think about a short description what's your problem and how do you think that it can be solved. Maybe someone had the same issue and/or knows some piece of code... | 20:06 |
thoralf | On which entrance task are you working on? | 20:06 |
dhruv13 | float comparison | 20:07 |
dhruv13 | it seems to fail an integration test | 20:08 |
dhruv13 | preprocessor_kernelpca_modular.py | 20:08 |
dhruv13 | i've been trying all day to reproduce the error, and i have a build going on which i think should do it | 20:08 |
dhruv13 | once that is done, i need to go to the tester file and change the way TParameter::equals is called from there | 20:09 |
-!- parijat [75c7d697@gateway/web/freenode/ip.117.199.214.151] has joined #shogun | 20:10 | |
thoralf | So you're waiting for travis to pass? | 20:10 |
-!- zxtx [~zv@ip-64-134-197-166.public.wayport.net] has quit [Ping timeout: 244 seconds] | 20:12 | |
-!- parijat [75c7d697@gateway/web/freenode/ip.117.199.214.151] has quit [Client Quit] | 20:13 | |
-!- sonne|osx [~sonne@e178248118.adsl.alicedsl.de] has joined #shogun | 20:13 | |
PirosB3 | hey guys, is it normal for my computer to take 2+ hours to compile Shogun?? :) | 20:13 |
-!- dhruv13_ [~dhruv13@84.64.13.210] has joined #shogun | 20:13 | |
thoralf | PirosB3: Depends on the number of cores and the interface you chose? | 20:14 |
PirosB3 | lol. I'm in a VM :( unfortunately I cannot get it to compile on Mac 10.7 :( | 20:14 |
dhruv13_ | thoralf: travis fails one particular case... it's issue #1888 by the way | 20:14 |
thoralf | PirosB3: Only enable the interface you need and set "make -j($cores+1)" | 20:14 |
dhruv13_ | :) | 20:15 |
-!- dhruv13 [~dhruv13@84.64.13.210] has quit [Ping timeout: 265 seconds] | 20:15 | |
PirosB3 | thoralf: I don't think I can exploit multicore | 20:15 |
PirosB3 | :( | 20:15 |
PirosB3 | Anyone on Mac 10.7 and can get it to build? | 20:16 |
thoralf | dhruv13_: Yes, I've seen your build. But currently you don't have a pending build, that's why I asked. | 20:16 |
@wiking | PirosB3: couldn't get it work with the macro hacks? | 20:17 |
PirosB3 | nope, it fails later on :( | 20:17 |
PirosB3 | it passes that phase, but fails on a c api | 20:17 |
dhruv13_ | thoralf: yeah, i can't seem to reproduce the error locally.... oh sorry, i was talking about my local build before... | 20:17 |
thoralf | dhruv13_: Okay, maybe we should check what you can do to build faster. 1. Do you use ccache? 2. Are you cleaning the complete build dir or are you reusing it? 3. What interfaces can be disabled? | 20:19 |
thoralf | Note on 2nd: If you change source files, you could (in theory) just hit make again in the build directory. | 20:20 |
dhruv13_ | thoralf: i don't use ccache, i'm reusing builds and i need the python debug interface | 20:20 |
thoralf | No need to delete it. | 20:20 |
dhruv13_ | thoralf: yeah, i'm not deleting. | 20:21 |
thoralf | Well. Why is it then taking so long? I guess it's swig? | 20:21 |
dhruv13_ | thoralf: problem is, i can't reproduce the error travis has on my machine | 20:21 |
dhruv13_ | thoralf: yeah, swig, and my crappy laptop | 20:21 |
dhruv13_ | it's pretty old actually | 20:23 |
dhruv13_ | 4 years + | 20:23 |
thoralf | You could try to use ccache to speed up your build, but it usually doesn't help on the swig part. | 20:23 |
dhruv13_ | will definitely try it... :) | 20:24 |
thoralf | s/usually// | 20:24 |
dhruv13_ | thoralf: i made changes to math.h, which is why it's probably taking long... | 20:30 |
thoralf | dhruv13_: How long did a build take before you changed it? | 20:32 |
thoralf | Do you think your fix slowed it down so much? | 20:33 |
dhruv13_ | thoralf: no, but math.h is used a lot right? and there hasn't been much change before and after | 20:34 |
dhruv13_ | thoralf: a build takes approx 20 mins | 20:34 |
dhruv13_ | thoralf: if i change math.h, and much less than 10 otherwise | 20:34 |
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 20:35 | |
thoralf | Sounds reasonable. | 20:35 |
-!- HeikoS [~heiko@nat-161-156.internal.eduroam.ucl.ac.uk] has joined #shogun | 20:49 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:50 | |
-!- dhruv13__ [~dhruv13@84.64.12.132] has joined #shogun | 20:51 | |
-!- dhruv13_ [~dhruv13@84.64.13.210] has quit [Ping timeout: 244 seconds] | 20:53 | |
lambday | HeikoS: trying to run the matlab code | 20:54 |
lambday | HeikoS: running few things on octave is tough, man! | 20:54 |
-!- dhruv13__ [~dhruv13@84.64.12.132] has quit [Ping timeout: 264 seconds] | 20:56 | |
@HeikoS | lambday: hey! | 20:57 |
@HeikoS | mtalab code? for the b-test paper? | 20:57 |
@HeikoS | never really looked at it, it should just serve as an idea | 20:57 |
lambday | HeikoS: yeah just for that | 20:57 |
lambday | well, don't need to run the code - | 20:58 |
lambday | would have been nice :-/ | 20:58 |
lambday | for the testcases | 20:58 |
@HeikoS | lambday: doesnt it work? | 20:58 |
@HeikoS | its *so* annoying that scientists still use Matlab to do stuff | 20:58 |
lambday | HeikoS: no man! internal.stat pkg doesn't exist - trying to download octave src then will configure with --enable-internal-stat :-/ | 20:59 |
lambday | grrr - I need tea! :( | 20:59 |
@wiking | besser82: custom command can depend on a target or only on file? | 20:59 |
@besser82 | wiking: can depend on target, and target can depend on fie ;) | 20:59 |
@besser82 | file | 21:00 |
@HeikoS | lambday: I just had a chai | 21:00 |
@HeikoS | lambday: was horrible European style ;) | 21:00 |
lambday | HeikoS: lol you still remember chai :P | 21:00 |
@HeikoS | lambday: sure :) | 21:01 |
lambday | HeikoS: man my Europe cycle tour :'( - don't know when I will do that :( | 21:01 |
@HeikoS | lambday: you should do a phd or so here, then you have loads of time | 21:01 |
lambday | HeikoS: last year was so envious when you went on that cycling tour - 95 kms a day | 21:02 |
lambday | man!!! | 21:02 |
@HeikoS | lambday: my father just asked me whether I wanted to do another one this summer :) | 21:02 |
lambday | HeikoS: hahahah! :D and??... | 21:02 |
@HeikoS | lambday: https://github.com/wojzaremba/btest/blob/master/btest.m | 21:03 |
@HeikoS | this doesnt work in octave? | 21:03 |
@HeikoS | lambday: will do it for sure :) | 21:03 |
lambday | HeikoS: yeah because of this line - https://github.com/wojzaremba/btest/blob/master/btest.m#L32 | 21:03 |
lambday | HeikoS: btw your interview was awesome! | 21:04 |
lambday | HeikoS: I can totally relate to paying rent with my gsoc money :D | 21:04 |
lambday | HeikoS: the only money that I have right now is that only :D | 21:04 |
@HeikoS | lambday: haha, great | 21:04 |
@HeikoS | lambday: in london you dont come far with it though | 21:04 |
@HeikoS | I needed to do two gsocs to survive for 6 months :) | 21:04 |
lambday | HeikoS: hehe :D | 21:05 |
@HeikoS | lambday: just edit that, its horrible style anyways | 21:05 |
@HeikoS | it justs sets parameters | 21:05 |
@HeikoS | lambday: you can just fix those | 21:05 |
lambday | HeikoS: checking.. | 21:06 |
lambday | this varargin is confusing | 21:07 |
lambday | checking | 21:07 |
@HeikoS | just fix the parameters, they are explained in the readme | 21:10 |
-!- zxtx [~zv@ip-64-134-197-166.public.wayport.net] has joined #shogun | 21:10 | |
-!- Saurabh7 [~Saurabh7@115.248.130.148] has joined #shogun | 21:12 | |
-!- thoralf [~thoralf@91-66-35-191-dynip.superkabel.de] has quit [Ping timeout: 244 seconds] | 21:16 | |
@wiking | HeikoS: http://en.cppreference.com/w/cpp/numeric/random | 21:22 |
@wiking | HeikoS: before u push more the boost story | 21:22 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 21:23 | |
@HeikoS | wiking: thats all nice but I think its only generating numbers | 21:23 |
@HeikoS | wiking: I am after log_pdf and cdf here | 21:23 |
@wiking | HeikoS: yeah i'm ust saying that mostly boost is used for it's PRNG | 21:24 |
@wiking | in stan | 21:24 |
@HeikoS | wiking: totally | 21:24 |
@wiking | i might be wrong | 21:24 |
@HeikoS | wiking: no the pdf functions are the important ones | 21:24 |
@wiking | but i dont know any other random distribution related code in boost | 21:24 |
@wiking | so if possible then maybe rather somehow replace the boost dependency with c++11 | 21:25 |
@wiking | if that's the only thing that is being used in boost | 21:25 |
@HeikoS | wiking: yeah I agree totally | 21:25 |
@HeikoS | wiking: what about this: | 21:25 |
@HeikoS | log_pdf is optional | 21:25 |
@HeikoS | if stan/boost is installed | 21:25 |
@HeikoS | and rng is done by c++ | 21:25 |
@wiking | lisitsyn: http://kep.index.hu/1/0/562/5625/56256/5625692_1b13051b81d5499cca6ce4083642bc82_wm.gif | 21:25 |
@HeikoS | ok gotta go, see you guys | 21:28 |
lambday | HeikoS: so soon? alright - I will probably start working on this by tomorrow | 21:29 |
lambday | HeikoS: understood what has to be done - need to make sure how to fit things inside the existing framework | 21:30 |
-!- pickle27 [c0008876@gateway/web/freenode/ip.192.0.136.118] has joined #shogun | 21:30 | |
pickle27 | sonney2k: around? | 21:30 |
lambday | oh before that, I should push the changes in the notebook | 21:31 |
-!- HeikoS [~heiko@nat-161-156.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 265 seconds] | 21:32 | |
@besser82 | wiking: was my hint helpful? | 21:37 |
@wiking | besser82: yeah i just need to now gather all the includes ... | 21:38 |
@besser82 | wiking: :D | 21:38 |
-!- zxtx [~zv@ip-64-134-197-166.public.wayport.net] has quit [Ping timeout: 240 seconds] | 21:40 | |
lambday | wtf :D http://9gag.tv/v/3705 | 22:02 |
lambday | I love her <3 http://9gag.tv/v/3715 | 22:07 |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 22:09 | |
sunil1337 | any entrance tasks for deep learning modules? | 22:10 |
pickle27 | just do any that you fancy! | 22:11 |
@wiking | sunil1337: you dont have to take a task that is related to your prefered project | 22:11 |
@wiking | entrance tasks are just to see how well you understand the structure of shogun itself | 22:12 |
sunil1337 | @wiking: Okay! :) | 22:12 |
@wiking | besser82: here? | 22:13 |
-!- sonne|osx [~sonne@e178248118.adsl.alicedsl.de] has quit [Quit: sonne|osx] | 22:20 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Quit: Page closed] | 22:23 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 22:27 | |
lisitsyn | wiking: yeah ;) | 22:32 |
-!- aroma [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has joined #shogun | 22:45 | |
kislay | do we have something in shogun to calculate the average image from a list of training images? i can work it out the conventional way using PIL but if there's anything in shogun that may help!! | 22:47 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 22:50 | |
shogun-notifier- | shogun: Viktor Gal :develop * a000824 / CMakeLists.txt,src/interfaces/matlab_static/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/a0008244202b68124dc3cf5658aecaadbbbdc986 | 22:50 |
shogun-notifier- | shogun: Fix Matlab static interface generation | 22:50 |
shogun-notifier- | shogun: Matlab static generation was not generated as stated in #1777 | 22:50 |
shogun-notifier- | shogun: This patch would not be possible without the help of @yage99 ! | 22:50 |
pickle27 | sonney2k: around? | 23:22 |
-!- thoralf [~thoralf@91-66-35-191-dynip.superkabel.de] has joined #shogun | 23:35 | |
thoralf | Hey. | 23:36 |
@wiking | lisitsyn: yo | 23:40 |
-!- pickle27 [c0008876@gateway/web/freenode/ip.192.0.136.118] has quit [Quit: Page closed] | 23:47 | |
--- Log closed Wed Mar 05 00:00:25 2014 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!