IRC logs of #shogun for Sunday, 2015-03-22

« Saturday, 2015-03-21 Index Monday, 2015-03-23 »
--- Log opened Sun Mar 22 00:00:09 2015
-!- curiousguy13 [~curiousgu@59.177.202.172] has quit [Ping timeout: 272 seconds]01:05
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun01:15
-!- curiousguy13 [~curiousgu@120.59.110.184] has joined #shogun01:18
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3_]03:14
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: leaving]04:14
-!- pickle27 [~pickle27@192-0-134-235.cpe.teksavvy.com] has quit [Remote host closed the connection]04:15
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun04:21
-!- mode/#shogun [+o wiking] by ChanServ04:21
-!- vortex_ape [~vortex_ap@182.64.177.121] has joined #shogun06:54
-!- curiousguy13 [~curiousgu@120.59.110.184] has quit [Ping timeout: 264 seconds]07:05
-!- curiousguy13 [~curiousgu@120.57.234.247] has joined #shogun07:18
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun09:19
-!- Nilesh [67157d50@gateway/web/freenode/ip.103.21.125.80] has joined #shogun09:46
NileshHello everyone09:46
-!- Nilesh is now known as Guest6912509:46
Guest69125I am Nilesh Kulkarni from Mumbai India09:47
Guest69125I needed help with working on LpBoost in shogun.09:47
Guest69125A C++ Example would be a great place to start09:52
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 264 seconds]10:47
-!- vortex_ape [~vortex_ap@182.64.177.121] has quit [Ping timeout: 246 seconds]10:57
-!- Guest69125 [67157d50@gateway/web/freenode/ip.103.21.125.80] has quit [Quit: Page closed]11:05
-!- vortex_ape [~vortex_ap@182.64.177.121] has joined #shogun11:16
-!- HeikoS [~heiko@0543fcaf.skybroadband.com] has joined #shogun11:49
-!- mode/#shogun [+o HeikoS] by ChanServ11:49
-!- vortex_ape [~vortex_ap@182.64.177.121] has quit [Ping timeout: 246 seconds]12:18
-!- vortex_ape [~vortex_ap@59.177.108.135] has joined #shogun12:30
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun12:51
shogun-notifier-shogun: jiaolong_cvc228 :develop * cb71681 / src/shogun/structure/GEMPLP.cpp: https://github.com/shogun-toolbox/shogun/commit/cb71681f8e4dc40d1b4e787a90abe7de73fc19bb12:51
shogun-notifier-shogun: fix defects reported by coverity scan12:51
shogun-notifier-shogun: Heiko Strathmann :develop * f4a2100 / src/shogun/structure/GEMPLP.cpp: https://github.com/shogun-toolbox/shogun/commit/f4a2100cdfb43cd278843cf771c316056ae75ac812:51
shogun-notifier-shogun: Merge pull request #2776 from Jiaolong/fix_defects_gemplp12:51
shogun-notifier-shogun:12:51
shogun-notifier-shogun: fix defects reported by coverity scan12:51
-!- nileshK [uid75191@gateway/web/irccloud.com/x-dqeidlpjndlozdle] has joined #shogun13:00
shogun-buildbotbuild #1012 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/1012  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, jiaolong_cvc228 <jiaolongxu@gmail.com>13:01
-!- travis-ci [~travis-ci@ec2-54-80-86-64.compute-1.amazonaws.com] has joined #shogun13:04
travis-ciit'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/5536217913:04
-!- travis-ci [~travis-ci@ec2-54-80-86-64.compute-1.amazonaws.com] has left #shogun []13:04
-!- curiousguy13 [~curiousgu@120.57.234.247] has quit [Ping timeout: 256 seconds]13:05
shogun-buildbotbuild #270 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/270  blamelist: jiaolong_cvc228 <jiaolongxu@gmail.com>13:09
shogun-buildbotbuild #271 of osx2 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/271  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:13
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun13:15
-!- curiousguy13 [~curiousgu@120.57.238.221] has joined #shogun13:17
-!- rajul [~rajul@node-wei.pool-125-24.dynamic.totbb.net] has joined #shogun14:08
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun14:32
-!- vortex_ape [~vortex_ap@59.177.108.135] has quit [Ping timeout: 246 seconds]15:04
-!- vortex_ape [~vortex_ap@182.64.69.25] has joined #shogun15:17
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 245 seconds]15:47
shogun-buildbotbuild #560 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/560  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, jiaolong_cvc228 <jiaolongxu@gmail.com>15:48
-!- HeikoS [~heiko@0543fcaf.skybroadband.com] has quit [Quit: Leaving.]15:51
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]15:51
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun17:08
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3_]17:39
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 250 seconds]17:41
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun17:42
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun17:49
nileshKHi I am facing a problem while running the ./classifier_lda example in shogun. Following is the output that I receive17:56
nileshK>>>> PrimalMosekSOSVM trained in    3.449917:57
nileshK>>>> BMRM trained in    5.572917:57
nileshK>>>> SGD trained in    0.781717:57
nileshK[WARN] In file /home/nilesh/BTP2/code2/shogun-3.1.1/src/shogun/multiclass/MulticlassStrategy.cpp line 45: MulticlassStrategy::CMulticlassStrategy(): register parameters!17:57
nileshK  0.516125  -0.226967  -0.528333   0.014236   0.421021  -0.353790   0.466174  -0.324728  -0.001395   0.298950   0.365580   0.115346   0.093586   0.312563  -0.510199   0.001264  -0.513814  -0.014525  -0.308745   0.17765117:57
nileshK  0.002279  -0.001445  -0.008696  -0.001968  -0.003419  -0.004643   0.007452  -0.006987  -0.004022   0.002983   0.001389  -0.000190   0.005914   0.004093  -0.004909  -0.005150  -0.006687  -0.006737  -0.003896   0.00097517:57
nileshK[ERROR] In file /home/nilesh/BTP2/code2/shogun-3.1.1/src/shogun/evaluation/StructuredAccuracy.cpp line 51: Unknown structured data type for evaluation17:57
nileshKterminate called after throwing an instance of 'shogun::ShogunException'17:57
nileshKAborted (core dumped)17:57
@wikingmmm17:59
@wikingwhy do you use still 3.1.1 version?17:59
@wikingwhy not 4.0.0?17:59
nileshKyeah, I was unable to compile18:04
nileshKthe 4.0.0 version18:04
nileshKI was getting some compilation with qsort something18:04
nileshKI got this version on the quickstart guide.18:05
nileshKI want a version of Shogun where I can easily get LpBoost working18:05
nileshKI have just been able to get the minimal part running18:05
-!- hargup [uid10222@gateway/web/irccloud.com/x-ohwbjurzeaqtvuph] has joined #shogun18:06
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 264 seconds]18:06
nileshKAnd the above errors w.r.t to running ./so_multiclas18:06
nileshKWhat is major change across 3.1.1 and 4.0.0?18:08
nileshKWould I be better off using it? In trying to get LpBoost working.18:08
nileshKAlso, if you pin point to me the exact commit I should use for compilation18:08
@wikingwhat's the error with qsort?18:08
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3_]18:09
nileshKThe file : examples/undocumented/libshogun/so_multilabel.cpp18:10
nileshKline no: 69 multilabel_sample.qsort();18:10
nileshKhas a compilation issue18:10
-!- pickle27 [~pickle27@192-0-134-235.cpe.teksavvy.com] has joined #shogun18:52
-!- vortex_ape [~vortex_ap@182.64.69.25] has quit [Ping timeout: 252 seconds]18:55
-!- vortex_ape [~vortex_ap@120.57.204.102] has joined #shogun18:56
-!- HeikoS [~heiko@0543fcaf.skybroadband.com] has joined #shogun19:01
-!- mode/#shogun [+o HeikoS] by ChanServ19:01
-!- curiousguy13 [~curiousgu@120.57.238.221] has quit [Ping timeout: 244 seconds]19:04
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun19:13
-!- curiousguy13 [~curiousgu@59.177.204.193] has joined #shogun19:17
-!- rajul [~rajul@node-wei.pool-125-24.dynamic.totbb.net] has quit [Quit: Leaving]19:22
-!- hargup [uid10222@gateway/web/irccloud.com/x-ohwbjurzeaqtvuph] has quit [Quit: Connection closed for inactivity]20:31
@HeikoSwiking: around?20:48
nileshKHeiko are you there?20:52
-!- vortex_ape [~vortex_ap@120.57.204.102] has quit [Ping timeout: 246 seconds]21:05
-!- vortex_ape [~vortex_ap@120.59.75.255] has joined #shogun21:06
@lisitsynHeikoS: hey21:08
@lisitsyncan I help with that stuff?21:08
@lisitsynneed some guidance to dive in21:08
@HeikoSlisitsyn: yes21:09
@HeikoSI need help! :=)21:09
@HeikoSWant to get some intro?21:09
@HeikoSthis is about hackinh sphinx21:09
@HeikoSI can tell you where I am21:09
@lisitsynyes please21:09
@HeikoShttps://groups.google.com/forum/#!topic/sphinx-users/1NmjM50DtMA21:09
@HeikoSso21:10
@HeikoSsphinx can do templates21:10
@HeikoSsee this one:21:10
@HeikoShttps://github.com/karlnapf/shogun_manual/commit/518a5204f687472156e12bca3628b2bd55f97c4c21:10
@HeikoSthe page.html is used to render the pages21:10
@lisitsynHeikoS: ok so you need a jinja21:10
@HeikoSas you can see, its quite easy to put the things in there21:10
@HeikoSsphinx is based in jinja21:10
@lisitsynto render a page that supports switching langs21:10
@HeikoSno sphinx can do all that21:11
@HeikoSif the {{body_python}} was defined21:11
@HeikoSthen we would be done21:11
@HeikoSbut one can only call {{body}} currently21:11
@lisitsynoh21:11
@HeikoSunderstand?21:11
@lisitsynwhere body_python comes from?21:11
@HeikoSI put it21:11
@HeikoSby hand21:11
@HeikoSthe remplate wont change its used for all sites21:11
@HeikoSso quite nice21:11
@HeikoSwe just need to create the corresponding body21:11
@HeikoSnow21:12
@lisitsynhmm I'd avoid that21:12
@lisitsynwhy not to have some 'bodies'21:12
@lisitsynand iterate over that?21:12
@HeikoSthat doesnt matter its a detail21:12
@HeikoSnot important now21:12
@HeikoSthis is just straight jinja2 templating21:12
@HeikoSthe problem is a different one21:12
@lisitsynseems important as while we have body_* we can't iterate over that21:12
@lisitsynok then21:12
@lisitsynI don't get the problem ;)21:13
@lisitsynwhat's it?21:13
@HeikoSi did not tell you yet21:13
@lisitsynoh ok :)21:13
@HeikoSdo you have sphinx installed?21:13
@lisitsyncan install that21:13
@HeikoSopen sphinx.builders.html.py21:13
@lisitsynwhere?21:14
@lisitsynhttps://github.com/karlnapf/shogun_manual21:14
@lisitsynhere?21:14
@HeikoSno21:15
@HeikoSjust plain sphinx21:15
@HeikoSneed to dive into its source code21:15
@lisitsynuhm21:15
@HeikoSas not properly documented21:15
@lisitsynI am not sure I have such a file21:15
@HeikoSit comes with spinx21:15
@lisitsynoh sorry21:16
@HeikoSgot it?21:16
@lisitsynI brewed wrong sphinx21:16
@HeikoSlet me know once you have it21:17
@lisitsynHeikoS: a sec21:17
@lisitsynHeikoS: okay I am here21:17
@HeikoSok21:18
@HeikoSso where the body is created is the function write_doc21:18
@HeikoSdef write_doc(self, docname, doctree):21:18
@HeikoSsee it?21:18
@HeikoSnow see here:21:19
@HeikoShttps://gist.github.com/karlnapf/d231d9d5587671645e7321:19
@HeikoSI added some functions to it21:19
@HeikoSthat manipulate the doctree, replace language specific parts, and then puts multiple "body" fields into the context dictionary21:20
@HeikoSso far so good21:20
@HeikoSbut the problem is: the doctree already contains the parsed sphinx .rst file21:20
@lisitsynokay I found write_doc21:20
@HeikoSthe function bascially takes the document tree representation, and produces the html21:21
@HeikoShandle_page call at the end adds it21:21
@HeikoSand ctx is a dictionary with the template arguments that will be replaced via jinja221:21
@HeikoSso ctx['body'] containts the html code that one can put in with {{body}} in the page.html21:21
@HeikoSget it?21:22
@lisitsynyeah probably21:22
@HeikoSok21:22
@lisitsynI need some explanation what .rst is21:22
@HeikoSah21:22
@HeikoSsee here:21:22
@HeikoShttps://gist.github.com/karlnapf/b97c4a07fcbd9959549621:23
@HeikoSits the way to write the page21:23
@HeikoSits where the code listings are included21:23
@HeikoSlisitsyn: ah you have to look at raw21:24
@HeikoShttps://gist.githubusercontent.com/karlnapf/b97c4a07fcbd99595496/raw/a85f1dc49dfd88614b892c45c9941b3e713df8b0/knn.rst21:24
@lisitsynlooks like markdown21:24
@lisitsynsimilar to21:24
@HeikoSits super similar21:24
@HeikoSits a superset though21:24
@lisitsynalright21:24
@lisitsynit it transformed to html21:24
@lisitsynit is*21:24
@lisitsynright?21:24
@HeikoSso now this literalinclude thing is transformed to syntax tree21:24
@HeikoSthen syntax tree is put into body21:24
@HeikoSI already know how to put in multiple body fields21:24
@HeikoSbut they are all based on the syntax tree which already contains the parsed python code21:25
@HeikoSso need access to the function that parses21:25
@HeikoSonce we have that, its done21:25
@HeikoSsomehow combine all21:25
@HeikoSso if you could check whether you can hack this thing to that the ctx dictionary in write_doc contains "body_python"21:26
@HeikoSthen all good21:26
@HeikoSits inserted into the template then21:26
@HeikoSand the page already works I have it running here locally21:26
@HeikoSwhich is cool as only a single sphinx make html produces all languages21:26
@HeikoSthen just need to do some finetuning, but not important for now21:26
@HeikoSimportant is to get the syntax tree for the different files21:26
@lisitsynok I think I understand21:27
@HeikoScheck this thing, it contains the way to add body_python etc21:27
@HeikoShttps://gist.github.com/karlnapf/d231d9d5587671645e7321:27
@lisitsynah so you patched sphinx to do that21:27
@HeikoSbut since the function receives the doctree as input (already parsed), it might not work this way21:28
@HeikoSone can also write sphinx extensions:21:28
@HeikoShttp://sphinx-doc.org/extensions.html#extensions21:28
@HeikoSI think that might be the cleaner way21:28
@HeikoSwe can then just define our own .rst command21:28
@HeikoSand then simply iterate21:28
@HeikoSlisitsyn: check this21:29
@HeikoShttp://sphinx-doc.org/extdev/tutorial.html21:29
@HeikoSmight be the way21:30
@lisitsynyeah looks like21:30
@lisitsynneed to glance over it a bit21:30
@HeikoSsomehow overload the :literalinclude21:30
@HeikoSnew command21:30
@HeikoS:shoguninclude21:30
@HeikoSwhere the python code calls the :literalinclude21:30
@HeikoSbut I think we also need to mofy the html builder to include multiple body fields21:31
@HeikoSlisitsyn: btw the code I sent in gist is not patching sphinx21:31
@HeikoSone can overload the class and tell sphinx to use it21:31
@HeikoSso thats fine21:31
@lisitsyncan we write :ourownliteralinclude?21:31
@HeikoSlisitsyn: yes21:31
@HeikoSlisitsyn: I guess :ourownliteralinclude would just concatenate multiple :literalinclude21:32
@lisitsynit seems that we shouldn't multiply boides21:32
@lisitsynbodies*21:32
@lisitsynbut just create a tab for each lang21:32
@lisitsynHeikoS: can this command be defined with jinja?21:33
@HeikoSwhich command21:33
@HeikoSlisitsyn: currently we have global tabs21:33
@lisitsynHeikoS: myliteralinclude21:34
@HeikoSwith a single body, tabs have to be local21:34
@HeikoSwhich is fine21:34
@lisitsynas wiking I like tabs just right to code more21:34
@HeikoSlisitsyn: I dont know, but I dont think thats the idea21:34
@HeikoSlisitsyn: well I was just going for easiest21:34
@HeikoSI think global is better in fact21:34
@HeikoSuser doesnt want to change locally for every snippet21:34
@HeikoSbut only once21:34
@lisitsynno21:34
@HeikoSbut thats super easy to do with template21:34
@lisitsynwe put it to parameter21:34
@HeikoSyeah21:34
@lisitsyn&lang=py21:35
@HeikoSbut thats all not important21:35
@HeikoSas its so easy to change21:35
@lisitsynonce selected21:35
@HeikoSmeat is in this problem here21:35
@lisitsynHeikoS: can we define a command :myinclude that is rendered by jinja template?21:35
@HeikoSso then your :shoguninclude would create the html tabs21:35
@HeikoSlisitsyn: I dont think it works that way21:35
@HeikoShave a look at21:35
@HeikoShttp://sphinx-doc.org/extdev/appapi.html#sphinx.application.Sphinx.add_directive21:36
@HeikoSlisitsyn: more this:21:36
@HeikoShttp://sphinx-doc.org/extdev/index.html#dev-extensions21:36
@HeikoSlisitsyn: it all works via nodes and python code21:37
@HeikoSso quite flexible21:37
@HeikoSone can iterate over all parameters21:37
@HeikoSwe could do:21:37
@HeikoS.. shogun:: examples/knn21:38
@HeikoS  :start-after: ![load_data]21:38
@HeikoS  :end-before: ![load_data]21:38
@lisitsynyes I think that's the idea21:38
@HeikoSwhich then iterates over all languages and calls the same code as in :literalinclude21:38
@HeikoSso why need jinja here?21:38
@lisitsynhmmm how do we define21:38
@lisitsynhow it is being rendered21:39
@lisitsynif it is just in body?21:39
@lisitsynHeikoS: in other words21:39
@lisitsynsay you have this shogun directive21:40
@lisitsynit loads all examples/knn.[lang]21:40
@lisitsynright?21:40
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 246 seconds]21:40
@HeikoSBuilderThe builder object (usually called builder) is an instance of a specific subclass of Builder. Each builder class knows how to convert the parsed documents into an output format, or otherwise process them (e.g. check external links).If you have the application object, the builder is available as app.builder.21:40
@HeikoSlisitsyn: yes21:40
@lisitsynok then where is the format is defined?21:40
@HeikoSthe builder thing would then do the tabbing21:40
@HeikoSBuilder21:40
@lisitsynohh21:40
@lisitsyncomplex thing21:40
@lisitsyn:D21:40
@HeikoSyes21:41
@lisitsynis there a jinja builder?21:41
@HeikoSlisitsyn: no need21:41
@lisitsynhow do we write html then?21:41
@HeikoSyou have python21:41
@HeikoSyou overload some methods21:41
@HeikoSwrite_doc method21:41
@HeikoSdont think thats a lot of work21:41
@HeikoSOR21:41
@HeikoSah no21:41
@HeikoSjust overload21:42
@lisitsynok theeen21:42
@HeikoSlisitsyn: wanna have a go on this?21:42
@lisitsynyes want to try21:42
@HeikoSIll have dinner now but would be back in an hour or so to help?21:42
@HeikoSI want to tackle this today! :)21:42
@lisitsynHeikoS: ok I'll try just now21:42
@HeikoShow long are you staying? :)21:42
@lisitsynHeikoS: for some random time21:43
@lisitsyngf could force me to talk21:43
@lisitsyn:D21:43
@lisitsynbut I can continue tomorrow anyway21:43
@HeikoSlisitsyn: ok21:44
@HeikoSwell ill have a look on the chat21:44
@HeikoSshare whatever you got21:44
@HeikoSIll work a bit more on it later21:44
@lisitsynHeikoS: I just need a way to generate that stuff21:44
@lisitsynis cloning your repo enough?21:44
@HeikoSno21:47
@HeikoSdo21:47
@HeikoSlisitsyn: sphinx-quickstart21:47
@HeikoSand then copy the knn.rst in21:47
@HeikoSIll add cource of the python and octave21:47
@lisitsynHeikoS: where to call sphinx-quickstart?21:49
@HeikoSlisitsyn: any dir21:49
@HeikoSand then populate with those files here:21:50
@HeikoShttps://gist.github.com/karlnapf/b97c4a07fcbd9959549621:50
@HeikoSthe html go to source/_templates21:50
@HeikoSthe .rst go to source21:50
@HeikoSand the examples to any folder of choice (matching the knn.rst)21:50
@lisitsynohh21:50
@HeikoSlisitsyn: good?21:50
@lisitsynokay21:50
@lisitsynyes I think I got it21:50
@HeikoSits just a minimal working sphinx config21:51
@HeikoSand the layout.html adds the bootstrap javascript21:51
@HeikoSand the page.html contains the layout for the tabs21:51
@HeikoSif you want to activate the bootstrap tabs, then you have to run21:51
@HeikoSpython -m SimpleHTTPServer 800021:52
@HeikoSand then navigate the the build/index.html21:52
@lisitsynok got it21:52
-!- PirosB3_ [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun22:12
@HeikoSlisitsyn: hi22:22
@HeikoSlisitsyn: any progress?22:22
@lisitsynHeikoS: just set up this thing22:23
@HeikoSlisitsyn: shall we maybe synch22:25
@HeikoSsync?22:25
@HeikoSpush to repo and then push incremental changes22:25
@HeikoSto see whats going on?22:25
@HeikoSor wanna stay in private?22:25
@lisitsynHeikoS: repo is better for sure22:25
@HeikoSlisitsyn: wanna push?22:27
@lisitsynHeikoS: yeah just a sec22:27
@HeikoScool22:27
@lisitsynHeikoS: hmm I see just tabs22:28
@lisitsynI guess something is b0rken22:28
@HeikoSthat means the body twas not found22:28
@HeikoSdo you have body_python ?22:28
@lisitsynah22:28
@HeikoSor just plain body?22:28
@lisitsynplain22:28
@HeikoSno that should do it22:29
@HeikoSI can also set up example and push?22:29
@HeikoSmaybe better?22:29
@HeikoSas I already got it working?22:29
@lisitsynHeikoS: yeah22:29
@lisitsynHeikoS: would be better22:29
@lisitsynHeikoS: you want all the doc be in sphinx?22:31
@HeikoSyes for now22:32
@HeikoSabout to push ...22:32
@lisitsynHeikoS: no I mean in general22:33
@HeikoSI dont care about this atm22:33
@HeikoSprobably in doc/manual/22:33
@HeikoSI want this to be seperate from doxygen22:33
@HeikoSand then populate/experiment22:33
@HeikoSand then maybe at some point use it for class docs also22:33
@lisitsynso doxygen stays for class reference?22:33
@HeikoSbut no need to do that now22:33
@HeikoScan change later22:33
@HeikoSpurpose is kind of different22:34
@lisitsynjust want to understand what you think22:34
@HeikoSbut it would be cool to have a uniform design22:34
@HeikoSah22:34
@HeikoSthis is an addition for now, so not replacing anything22:34
@HeikoSBut I want to move *all* doxygen documentation that is not md-file or class docs to sphinx22:34
@lisitsynI see22:34
@HeikoSmd files come from doc submodule22:35
@HeikoSand class docs are fine for now22:35
@HeikoSserve purpose22:35
@HeikoSwe should also add a :shogunclass directive22:35
@HeikoSthat links against class docs for a class22:35
@HeikoSrather than using hard-links in the examples22:35
@HeikoSbut let me push22:35
@HeikoSlisitsyn: ok22:39
@HeikoSclone it22:39
@HeikoSgo to sphinx dir22:39
@HeikoStype make html22:39
@HeikoSand then python -m SimpleHTTPServer 800022:39
@HeikoSthen you will see the tabs22:39
@lisitsyncool thanks22:40
@HeikoSlisitsyn: for now I left the global tabs22:41
@HeikoSbut they all just include {{body}}22:41
@HeikoSsee page.html22:41
@HeikoSlets try to write this extension together22:41
@HeikoSmaybe for now clone the :literalinclude?22:42
@HeikoSand then modify?22:42
@lisitsynHeikoS: ok it works22:42
@lisitsynHeikoS: where literalinclude is defined?22:42
@HeikoShis is what you can do in an extension: First, you can add new builders to support new output formats or actions on the parsed documents. Then, it is possible to register custom reStructuredText roles and directives, extending the markup. And finally, there are so-called "hook points" at strategic places throughout the build process, where an extension can register a hook and run specialized code.22:43
@HeikoShttp://sphinx-doc.org/extdev/index.html#dev-extensions22:43
@HeikoSlets do this :22:44
@HeikoShttp://sphinx-doc.org/extdev/tutorial.html22:44
@HeikoSso we need this setup function first22:45
-!- vortex_ape [~vortex_ap@120.59.75.255] has quit [Ping timeout: 264 seconds]22:45
@lisitsynhmm22:45
@lisitsynlooks simple lets try22:45
@HeikoSwhere does this file go?22:46
@HeikoSjust python path?22:46
@lisitsynHeikoS: I am not sure22:46
@lisitsynhave the same question22:46
@lisitsynHeikoS: I think source22:46
@lisitsynconf.py is already here22:46
@HeikoSkk22:46
@lisitsynHeikoS: lets just copy paste22:46
@lisitsynHeikoS: I added this thing22:47
@lisitsynnothing happened22:47
@HeikoSyeah22:48
@HeikoSI think this todo.py adds directives22:48
@HeikoSin the conf.py we might be able to add the extension22:48
@lisitsynHeikoS: I am not sure it is loaded22:48
@HeikoSextensions = [22:49
@HeikoS    'sphinx.ext.mathjax',22:49
@HeikoS    'todo'22:49
@HeikoS]22:49
@HeikoSin conf.py22:49
@lisitsynahhh22:49
@lisitsynI see22:49
@lisitsynno module named todo22:49
@HeikoSmove to makefile dir22:49
@HeikoSthen youll get another error ;)22:49
@HeikoSgot it?22:50
@HeikoStodolist not defined?22:50
@HeikoSthey say in the tutorial: We need to create the two node classes todo and todolist later.22:50
@lisitsynah yeah22:50
@lisitsynHeikoS: then it just catches modules from currentpath + PYTHONPATH22:51
@HeikoSyes22:51
@HeikoSbut thats fine22:51
@lisitsynyeah more or less22:51
@HeikoSthe thing can be in makefile dir22:51
@HeikoSor somewhere else22:51
@lisitsynwe just modify PYTHONPATH22:51
@HeikoSyep22:51
@HeikoSok so next step is to define the classes22:51
@lisitsynHeikoS: pythonpath again I guess22:52
@HeikoSI copied this to the same file22:52
@HeikoSto the beginning22:52
@lisitsynyeah that' s good22:52
@HeikoSlisitsyn: but also doesnt work yet22:52
@lisitsynHeikoS: I put all the things below22:54
@lisitsynworks22:54
@HeikoSlisitsyn: can you past your file in some gist?22:54
@HeikoSlisitsyn:  I get22:55
@HeikoS  File "/home/heiko/git/shogun/shogun_manual/sphinx/todo.py", line 4, in setup22:55
@HeikoS    app.add_node(todolist)22:55
@HeikoSNameError: global name 'todolist' is not defined22:55
@lisitsynHeikoS: https://gist.github.com/lisitsyn/a885cd193a4b1b7fae7922:55
@HeikoSdoesnt work22:56
@HeikoSsigh22:56
@HeikoSah22:56
@HeikoSmy mistake22:56
@HeikoSworks22:56
@lisitsynHeikoS: ok then it is quite simple22:57
@lisitsynfollowing this example we can add shoguninclude thing22:57
@HeikoSdid you try to use it?22:57
@HeikoSill try22:57
@lisitsynnot yet22:57
@HeikoSdont get a todo list22:59
@lisitsynyeah me too22:59
@HeikoSbut dont know the syntax of the directive23:00
@HeikoS   .. todo:: asdasd23:00
@HeikoS   .. todo:: asdasd223:00
@HeikoS   .. todolist23:00
@HeikoSI suggest we try to add tabs around the todolist :)23:01
@lisitsynHeikoS: :todo_include_todos: true23:02
@lisitsynaccording to http://sphinx-doc.org/ext/todo.html?highlight=todo#module-sphinx.ext.todo23:02
@HeikoSconfig?23:02
@lisitsynyeah I guess23:02
@lisitsynhmm still not23:03
@HeikoSI get lots of non existing references when building23:05
@lisitsynHeikoS: ha!23:05
@lisitsynworked23:05
@HeikoSpush!23:06
@lisitsynok23:06
@HeikoSand then we can add the tabs to the todo list23:06
@lisitsynHeikoS: a second23:08
@HeikoSlisitsyn: you have push rights23:09
@lisitsynHeikoS: done23:10
@lisitsyntry that23:10
@lisitsynHeikoS: use make preview23:10
@lisitsynthat's just make html and python -m thing23:10
@HeikoSI see23:11
@HeikoSpulled23:11
@lisitsynok that's pretty straightforward23:11
@lisitsynHeikoS: let me take care of this but tomorrow morning23:11
@HeikoSwanna stop?23:12
@lisitsynHeikoS: gf is waiting sorry :)23:12
@HeikoSno worries23:12
@HeikoSI might play a bit23:12
@HeikoShave a good night!23:12
@lisitsynHeikoS: I think that's the way to go23:13
@lisitsynI guess sphinx people would suggest the same thing23:13
@lisitsynok see you tomorrow23:13
@HeikoSsee you!23:14
@lisitsynHeikoS: still here?23:25
@lisitsynhttps://www.kaggle.com/c/otto-group-product-classification-challenge lets beat people on this small data23:25
@lisitsynHeikoS: your GP expertise should rock here23:26
@HeikoShaha23:26
@HeikoSyes23:26
@HeikoSwould love to23:27
@HeikoSbut time issues as usual23:27
@lisitsynHeikoS: be my shogun consultant then23:27
@HeikoSsure :)23:28
@lisitsynI love small data23:28
@lisitsynbuilding infinite dimensional models is like touching the god23:28
@lisitsynso I start with kernel svm23:28
@lisitsynHeikoS: what kernel rocks for 0,1,2,...~40 counters?23:29
@HeikoShehe23:30
@HeikoSuh thats discrete23:30
@HeikoSuse Gaussian23:30
@HeikoSis it ordered?23:30
@HeikoSthen Gaussian is fine23:30
@lisitsynHeikoS: 0,0,0,0,0,0,0,4,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0,0,0,0,2,0,0,0,2,0,0,0,0,0,0,0,4,0,0,0,0,0,0,0,0,0,0,16,0,0,3,0,0,0,0,23,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,5,1,0,0,0,0,0,0,Class_923:31
@lisitsyntypisch vector23:31
@HeikoSuse gaussian as a baseline23:32
@HeikoSwill do good23:32
@HeikoSto me it seems sparse?23:33
@HeikoSactually23:33
@HeikoStry a linear model23:33
@HeikoSliblinear23:33
@HeikoSshould do as well23:33
@HeikoSthen compare to Gaussian23:33
@HeikoSthen compare to random forest23:33
@HeikoSwe need this supervised learning framework23:33
@HeikoSfor shogun23:33
@HeikoSwhere all models are fitted, parameters learned automagically23:33
@HeikoSand then boosting or so learns the weights23:33
@lisitsynHeikoS: the more kaggle tries I do the better I understand how should it look liek23:33
@lisitsynHeikoS: I bet linear would not win here23:34
@HeikoStry23:34
@HeikoSbaseline for sure23:34
@lisitsyn# features <<< # vectors23:34
@lisitsyn38 features23:34
@lisitsyn60K train examples23:35
@lisitsynHeikoS: obviously more complex thing is possible23:35
@lisitsynHeikoS: would this diffusion kernel rock here?23:35
@lisitsynHeikoS: I have been thinking about it as a panacea for such discrete features haha23:36
@lisitsynthey look very math :D23:36
@HeikoSthe diffusion kernel is for unordered discrete data23:37
@HeikoSif you have an ordering its fine to embed into real space23:37
@lisitsynah23:37
@lisitsynI see23:37
@HeikoSbut shogun doesnt have categorial feature representation that is good23:37
@HeikoSif we had23:37
@HeikoSI could implement the diffusion kernel23:37
@HeikoSwhich is very good23:37
@lisitsyndid you try it on any task?23:37
@HeikoSbetter then 1-in-n encoding and then gaussian23:37
@HeikoSyes23:37
@HeikoSuci dataset23:37
@HeikoSmushrooms23:37
@lisitsynHeikoS: do you think there is possible linear approximation ?23:38
@lisitsynlike random gaussian features or HKM23:38
@HeikoSyes for sure23:38
@HeikoSthough not that clear23:38
@lisitsynis it known already?23:38
@HeikoSdont know23:38
@lisitsynlets formulate if not23:39
@lisitsynand win all the kaggles23:39
@lisitsynHeikoS: I'd google for that tomorrow23:40
@lisitsynok I think I am in danger if I keep using my notebook now23:40
@lisitsynsee you tomorrow :)23:40
@HeikoShehe23:41
@HeikoSgood luck23:41
--- Log closed Mon Mar 23 00:00:11 2015
« Saturday, 2015-03-21 Index Monday, 2015-03-23 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!