IRC logs of #shogun for Wednesday, 2013-05-08

« Tuesday, 2013-05-07 Index Thursday, 2013-05-09 »
--- Log opened Wed May 08 00:00:35 2013
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]00:11
pickle27iglesiasg: here is a new pastebin http://pastebin.com/KsR7kLw600:44
pickle27iglesiasg: true, I meant the example runs00:45
iglesiasgpickle27: the pastebin is still not loading here for some reason00:46
pickle27iglesiasg: out of curiosity I just ran valgrind on classifier_qda and I got the same unrecognized instruction line00:47
pickle27iglesiasg: strange00:47
@sonney2kiglesiasg, it is loading here00:48
@sonney2ktry a different browser00:48
@sonney2kpickle27, are you sure you compile w/o optimization?00:49
iglesiasgsonney2k: I should get a proper internet connection rather :S00:49
@sonney2kiglesiasg, or that00:49
pickle27yes I've compiled without optimization and with debugging00:49
iglesiasgsonney2k: my mom has not got wifi at home so I get connected here with a 3G dongle00:49
pickle27ooooj00:50
pickle27hold up00:50
@sonney2kpickle27, and you make ; make installed?00:50
pickle27I just realized that I have not make installed00:50
pickle27wow00:50
@sonney2kand then also compiled the example again?00:50
@sonney2khaha00:50
iglesiasgpickle27: hehe00:50
@sonney2kgotcha00:50
pickle27literally realized the second before you typed that00:50
iglesiasgpickle27: alias make=make; sudo make install00:50
iglesiasgor sth like that, please :P00:51
@sonney2kwhy root?00:51
@sonney2kI usually install somewhere where that is not needed00:52
iglesiasgsonney2k: just got the habit to do it for python interface00:52
pickle27okay now I have something00:52
@sonney2kpickle27, show us00:52
@sonney2kwell me00:52
pickle27http://pastebin.com/Ek8PW8cZ00:52
iglesiasgsonney2k: so is it actually normal that after compiling with optimizations leaks may appear??00:52
@sonney2kspanish can't afford :P00:53
iglesiasg:(00:53
iglesiasgpickle27: maybe if you run it without -v we still got useful info and my connection can handle that :)00:54
pickle27iglesiasg: http://pastebin.com/EwEKypN000:54
@sonney2kiglesiasg, nope but valgrind may crash00:55
iglesiasgsonney2k: Isee00:55
@sonney2kpickle27, is S empty?00:55
@sonney2kpickle27, maybe doing00:56
pickle27sonney2k: sorry new to using valgring where are you reading00:57
@sonney2kwhile ( rank < m_dim) && S[rank] > m_tolerance)00:57
@sonney2kwould fix it?00:57
pickle27oh in code00:57
@sonney2kwell line 34100:57
@sonney2kin MCLDA.cpp00:57
@sonney2kyou access S[rank]00:57
pickle27yeah I see00:58
pickle27S shouldn't be empty it should have been filled by wrap_dgesvd00:58
@sonney2kbtw no idea who merged this but please use { } brackets around for loop's with >1 line00:58
@sonney2kand indention seems to be broken00:58
@sonney2kyou have spaces & tabs mixed00:58
@sonney2kwe use tabs only00:59
@sonney2ksize 400:59
@sonney2kand you can write00:59
@sonney2kinstead of SGVector< float64_t > S  = SGVector< float64_t >(m_dim);00:59
@sonney2kSGVector< float64_t > S(m_dim);00:59
pickle27yeah I know my tabs are off, is there a better editor for fixing that? gedit displays it fine00:59
iglesiasgpickle27: vim is better than gedit indeed :)01:00
@sonney2kpickle27, well open gedit's preferences and go to editor -> set tab width to 401:00
iglesiasgbut gedit should be just fine for the tabs honestly01:00
pickle27yeah I have done that but it makes it appear to be formatted okay when it isn't01:00
@sonney2kthen maybe enable the plugin01:01
iglesiasgpickle27: is there an option to show whitespace characters?01:01
@sonney2k'draw spaces'01:01
pickle27sonney2k: and you prefer braces around all loops?01:02
@sonney2kpickle27, when they are > 1 line yes01:04
@sonney2kin particular with the current indention massacre it is hard to read01:04
@sonney2kthat indention == part of the language thing of python is really sth I miss in other $LANGs01:05
pickle27yeah Im working on that sorry01:05
@sonney2kpickle27, btw in python spaces are now mandatory or do they allow tabs?01:05
pickle27I think they allow tabs01:06
@sonney2kin 2.x they did01:06
@sonney2kbut we all know how dangerous this is01:06
-!- Ingvaras [522c194e@gateway/web/freenode/ip.82.44.25.78] has quit [Ping timeout: 245 seconds]01:06
iglesiasgsonney2k: why do you like more tabs than spaces btw?01:07
-!- mode/#shogun [+o iglesiasg] by ChanServ01:08
@sonney2kiglesiasg, nowadays it is probably no longer reasonable01:09
@sonney2kbut back in the days (recall I am old ;)01:09
@sonney2ktab == 1 byte01:09
@sonney2keditors didn't have smart tabs01:09
@sonney2kreplacing spaces with tabs etc01:09
@iglesiasgoh interesting01:10
@sonney2kit's like linux kernel01:10
@sonney2kthey use tabs too01:10
@sonney2kthough the kernel is a few more decades older than me01:11
@iglesiasgheh01:11
pickle27hey I think it might be good now!01:12
@sonney2kpickle27, what did you change?01:13
@iglesiasgboth indentantion and the bug?01:13
pickle27http://pastebin.com/M7HaiUpE01:13
pickle27yes both01:13
@sonney2kuahhah!01:13
pickle27thats the entire output of valgrind01:13
@sonney2kpickle27, so was it what I suggested?01:13
@sonney2kor what?01:13
pickle27got to hand it to you that S was a good catch01:13
pickle27I also figured out my gedit problem I think01:14
pickle27you need to restart it after changing from spaces to tabs01:14
@sonney2kI have wasted^H^H^H^Hspent too much time of my life hunting bugs01:14
@sonney2kso I know a good part of their species01:15
@iglesiasgyou have gotten a 6th sense indeed01:15
pickle27sonney2k: ah but the thrill when you catch one!01:15
@sonney2kand I am collecting them as trophies01:15
@sonney2kbut only the very rare hard to catch ones nowadays01:15
@sonney2kpickle27, so keep that one for yourself01:15
@sonney2kI prefer multithreading bugs these days01:16
@sonney2kthey just shiiiiinnne01:16
pickle27sonney2k: :)01:16
@sonney2koh dammn I need my medication01:16
@sonney2kand sleep01:16
@sonney2kgood night everyone!01:16
@sonney2kand hurray 8 slots!01:17
@iglesiasgnight01:17
pickle27night01:17
@iglesiasgsonney2k: let's see if we get any more with the re-allocation01:17
@iglesiasgproviding that last year we gave one, we have the karma with us :)01:18
pickle27alright I'm out for the night too now01:19
pickle27later!01:19
@iglesiasgpickle27: talking about funny bugs01:19
@iglesiasghttps://github.com/shogun-toolbox/shogun/commit/2234df4636c1e6061fcc19209fe279850efa735401:19
@iglesiasgsee you!01:19
-!- iglesiasg [d58f3213@gateway/web/freenode/ip.213.143.50.19] has quit [Quit: Page closed]01:27
-!- foulwall [~foulwall@2001:da8:215:503:1943:d77b:b38a:a922] has joined #shogun02:08
-!- madhura [~madhura@59.95.237.64] has quit [Quit: Leaving]02:10
-!- shogun-toolbox [~shogun@7nn.de] has quit [Ping timeout: 264 seconds]02:20
--- Log closed Wed May 08 02:20:30 2013
--- Log opened Wed May 08 02:20:39 2013
-!- shogun-t1olbox [~shogun@7nn.de] has joined #shogun02:20
-!- Irssi: #shogun: Total of 17 nicks [3 ops, 0 halfops, 0 voices, 14 normal]02:20
-!- Irssi: Join to #shogun was synced in 8 secs02:20
-!- shogun-toolbox [~shogun@7nn.de] has quit [Ping timeout: 264 seconds]02:20
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 252 seconds]02:46
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun03:32
-!- poojits [75d35d4a@gateway/web/freenode/ip.117.211.93.74] has quit [Ping timeout: 245 seconds]03:41
-!- zxtx [~zv@68.125.89.66] has joined #shogun03:59
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun04:39
-!- nube [~rho@49.244.128.165] has quit [Ping timeout: 245 seconds]04:58
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]05:04
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has joined #shogun05:24
-!- foulwall [~foulwall@2001:da8:215:503:1943:d77b:b38a:a922] has quit [Remote host closed the connection]05:27
-!- lisitsyn [~blackburn@188.168.3.6] has quit [Quit: Leaving.]05:47
-!- foulwall [~foulwall@2001:da8:215:503:c16d:6025:8431:c642] has joined #shogun05:53
-!- zxtx [~zv@68.125.89.66] has quit [Ping timeout: 248 seconds]06:55
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun07:13
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has quit [Ping timeout: 245 seconds]07:18
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds]07:24
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun07:29
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]07:44
foulwallHi sonney2k?08:00
@sonney2kfoulwall, morning08:21
foulwallsonney2k: sorry for my absence, I left for my grandpa's funeral. I'll start my work today.08:22
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun08:23
foulwallsonney2k: I'm trying to do the kernel matrix visualization work.08:24
@sonney2kfoulwall, my condolences08:26
@sonney2kfoulwall, just for the record - it is no problem if you are away but please tell us upfront to avoid misunderstandings in the future08:26
foulwallsonney2k: I'll let you know when I am away! sorry for not telling.08:28
@sonney2kso I hope you recovered from the funeral.. losing someone is always #%$!08:28
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has joined #shogun08:28
shogun-buildbotbuild #926 of bsd1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/926  blamelist: Kevin Hughes <kevin@kevin-M3920.(none)>08:28
foulwallsonney2k: so anything will be all right. I return to work.08:29
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun08:30
deerishihi everyone!!08:30
deerishineone der?08:30
deerishii wanna work on Cross-Product for lists of kernels bug08:31
deerishii have some difficulty understanding the problem statement08:31
@sonney2kdeerishi, I think van51 worked on that already08:32
shogun-buildbotbuild #927 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/92708:42
shogun-buildbotbuild #761 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/761  blamelist: Soeren Sonnenburg <sonne@debian.org>08:44
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun09:14
-!- mode/#shogun [+o lisitsyn] by ChanServ09:14
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun09:23
-!- mode/#shogun [+o iglesiasg] by ChanServ09:26
@lisitsynfoulwall: condolences09:43
-!- foulwall [~foulwall@2001:da8:215:503:c16d:6025:8431:c642] has quit [Remote host closed the connection]09:45
@sonney2klisitsyn, I am considering to switch the mailing list manager for shogun toolbox.org09:55
@lisitsynsonney2k: why?09:55
@sonney2kI am currently setting up a new US based server09:55
@sonney2kthe mailing list manager is very slim & neat admittedly09:56
@sonney2kbut it only works well with qmail09:56
@sonney2kand qmail is not really maintained09:56
@sonney2kso I am going for the default -> exim09:56
@sonney2kI hate mailman for doing .mbox files09:57
@sonney2kI mean one big mbox file growing and growing09:57
@sonney2knot one file / mail09:57
@sonney2kthat is so stupid09:57
@lisitsynI realize I have absolutely no clue what is you are talking about :D09:58
@sonney2know when you do backups you backup the gigabyte big file09:58
@sonney2kif someone just sent a *single* new mail to the list09:58
-!- foulwall [~foulwall@2001:da8:215:503:c47e:67cf:6bd7:93a9] has joined #shogun10:14
@sonney2klisitsyn, I just did the switch10:28
@sonney2ktrouble ahead...10:29
@lisitsynsonney2k: what kind of?10:29
@sonney2kmailinglist berserk / mail not working10:29
@sonney2kfoulwall, the good news is that we will have some separate virtual server to play with for the demos :D10:30
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has joined #shogun10:30
foulwallsonney2k: That's great110:30
-!- nube [~rho@116.90.239.3] has joined #shogun10:34
-!- HeikoS [~heiko@nat-179-224.internal.eduroam.ucl.ac.uk] has joined #shogun10:47
-!- mode/#shogun [+o HeikoS] by ChanServ10:47
Yanglittleexcuse me i can't load multiclasslabels from shogun.Features under ubuntu 13.0410:48
Yanglittlebut others can load such as combinedfeatures10:49
@lisitsynYanglittle: hey. what do you mean by load?10:51
@lisitsynHeikoS: hey10:51
@HeikoShi!°10:51
@HeikoSfinally back :)10:51
@lisitsynwe haven't seen you for a while ;)10:51
Yanglittlei mean import10:51
@HeikoSthanks for doing all the replies10:51
@lisitsynHeikoS: have you seen we've got 8 slots?10:51
foulwallmaybe eigen3 dependency?10:51
@HeikoSlisitsyn: yes, told you I was in Germany and had only minor internet access10:51
@HeikoSlisitsyn: I have seen it10:52
@lisitsynI remember10:52
@HeikoSthis is pretty much what we wanted right?10:52
@HeikoSYanglittle: might be eigen or another dependency indeed10:52
@lisitsynHeikoS: we should have some decision on two unlucky we left out10:52
@lisitsynI don't think it is eigen dependency10:52
@lisitsynbut let me check10:52
@lisitsynno, no dependency here10:53
@HeikoSlisitsyn: could you explain?10:54
@HeikoSlisitsyn: why did you already give feedback btw? isnt that supposed to happen after its fixed?10:54
Yanglittleeigen support  is yes.10:54
@lisitsynHeikoS: just to push students to be more active10:54
@lisitsynthat's ok10:54
@sonney2kHeikoS, what?10:55
@sonney2khey there10:55
@sonney2kHeikoS, we give feedback to figure out whom to accept10:56
@sonney2kif everyone just sits there and wait then we will never find out who fits best10:56
@HeikoSsonney2k: hi!10:57
@HeikoSsonney2k: sergey told me, I agree with that, just wondered when I first saw it10:58
@sonney2kwell last night :D11:00
@HeikoSsonney2k: I just read stuff on my phone, no point in writing on it :D11:00
@sonney2kHeikoS, I was thinking a bit about our current tests.11:02
@HeikoSsonney2k:  yes?11:02
@sonney2kHeikoS, what we do is serialize and unserialize11:02
@sonney2kthen compare if results are exactly the same11:03
@HeikoSyes11:03
@sonney2know we have the issue w/ gaussian_process_regression11:03
@sonney2kthat 3 out of 100 labels deviate by 1e-1511:03
@HeikoSyep11:03
@HeikoSthat should be fine though11:04
@HeikoSjust update the test11:04
@HeikoSannyoing11:04
@sonney2kno11:04
@HeikoSbut as you said, its a one to one comparison11:04
@sonney2kit works fine everywhere except on travis11:04
@HeikoSI like this regarding the states of the instances11:04
@HeikoSah11:04
@HeikoSwhat about a comparison class then?11:04
@HeikoSthat compares all parameters to be close to each other?11:04
@sonney2kso lets assume that this result is OK (which is not clear to me btw...)11:04
@sonney2kyes exactly11:05
@sonney2kthat is what I wanted to ask11:05
@sonney2kwe should have a class / method that utilizes the parameter framework11:05
@lisitsynsonney2k: we could check if their virtual machine image is available somewhere11:05
@sonney2kand then compares parameters with a certain precision11:05
-!- gsomix [~gsomix@188.168.3.6] has quit [Ping timeout: 252 seconds]11:05
@sonney2klisitsyn, I suspect that it is due to numpy 1.6.111:06
@lisitsynsonney2k: random?11:06
@sonney2kwell everything is the same (6 or seven results returned)11:06
@sonney2kbut labels in 3 things?!11:06
@sonney2kor it could be multithreading11:06
@lisitsynhah yes totally crazy11:06
@sonney2kI mean the mean of the GP and everythign11:07
@HeikoSsonney2k: yes that would be nice to have11:07
@HeikoSworks fine for all numerical stuff11:07
@HeikoSand for SGOnbjects, we do recursion11:07
@sonney2kHeikoS, we know the type right?11:07
@sonney2kyeah TParameter11:07
@HeikoSyes11:07
@HeikoSsome work though11:07
@HeikoSbut I can do that11:08
@HeikoSat some near point11:08
@sonney2kso we could just have a SGObject equals() method11:08
Yanglittleexcuse me, are there any methods to solve the problem?11:08
@sonney2kthat takes an SGObject11:08
@lisitsynYanglittle: oh sorry11:08
@HeikoSsonney2k: yes11:08
@lisitsynYanglittle: so this fails:11:08
@sonney2kthen works the same way as list_parameters11:08
@lisitsynfrom modshogun import MulticlassLabels11:08
@lisitsynright?11:08
@HeikoSsonney2k I dont see any big obstacles with that11:08
@sonney2kHeikoS, if we had this - we could even extend the testsuite to other languages11:09
@sonney2kso we no longer need pickle11:09
@sonney2kbut can directly compare shogun objects created in python_modular with java_modular11:09
Yanglittlefrom modshogun import MulticlassLabels Traceback (innermost last):   File "<stdin>", line 1, in <module> ImportError: cannot import name MulticlassLabels11:09
@sonney2ketc11:09
@HeikoSsonney2k, yes, that sounds good, although we pay with a  more complicated thing that might be buggy11:09
@HeikoSbut still I like that11:09
@HeikoSI would do it after fixing the migration stuff however11:09
@HeikoSbecause currently we cannot alter SGObjects without breaking the file format11:10
@sonney2kwell a unit test would ensure that comparing stuff works11:10
@HeikoSand the tests11:10
@HeikoSsonney2k, yes11:10
@sonney2kboth is critical11:10
@sonney2kwe can hardly afford the buildbot failing etc11:10
@HeikoSyes11:11
@HeikoSso the equals method should be easy and nice to write11:11
@HeikoSthe migration stuff is a nightmare to solve11:11
@HeikoSno idea where the problem is etc11:11
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 264 seconds]11:11
@lisitsynYanglittle: is that latest version? I see no reason to fail there actually11:12
Yanglittleyes the latest version,11:13
@sonney2kHeikoS, shouldn't a small unit test for the migration stuff expose that already?11:15
@sonney2kI mean if it is small - one can debug it nicely11:15
-!- nube [~rho@116.90.239.3] has joined #shogun11:16
sijinlisitsyn: if Travis build fails, how to figure out where is the problem? it only tells that one file conflicts11:18
@HeikoSsonney2k: yes maybe11:18
@HeikoSits just that the code is so complicated that I have trouble understanding it11:19
@HeikoSbut in principle you are right11:19
@lisitsynsijin: conflicts? like?11:19
@HeikoSsonney2k: but you are right, this is the way to go11:19
@HeikoSstart with the very fundamental parts and work up11:20
sijinI got this https://travis-ci.org/shogun-toolbox/shogun/builds/6634337 several days ago?11:20
@HeikoSsome parts cannot be checked via unit tests though11:20
@HeikoSsince you need classes that alter etc11:20
@HeikoSbetter to it in an example11:20
@lisitsynsijin: oh that's not caused by your changes I think11:21
-!- nube1 [~rho@116.90.239.3] has joined #shogun11:21
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 276 seconds]11:22
sijinlisitsyn: so what I can do for it? I didn't find test for Kpca11:22
@sonney2ksijin, write one!11:24
@lisitsynsijin: yeah we have no test11:24
@lisitsynsijin: I just glanced over your PR - could you please fix tabs/spaces? it is wavy sometimes11:24
sijinsonney2k: ok. What need to be tested? I mean there is no ground truth for kpca, compare the result with another package?11:25
sijinlisitsyn: Fine, I will do that.11:26
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has quit [Quit: Page closed]11:46
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun11:46
shogun-notifier-shogun: hushell :develop * d39547b / / (11 files): https://github.com/shogun-toolbox/shogun/commit/d39547b50c0e97c690d228c9f5f30c8c408c71ca11:46
shogun-notifier-shogun: multiclass heuristics for probabilistic outputs estimations11:46
shogun-notifier-shogun: hushell :develop * 2bb00e1 / / (6 files): https://github.com/shogun-toolbox/shogun/commit/2bb00e1745c7192c8c6449bdd784b574ff970ae911:46
shogun-notifier-shogun: added unit-test11:46
shogun-notifier-shogun: hushell :develop * a3faafe / src/shogun/multiclass/ (5 files): https://github.com/shogun-toolbox/shogun/commit/a3faafeb2fc5110288fa68c39c775b5218be348411:46
shogun-notifier-shogun: dropped SGVector reference11:46
shogun-notifier-shogun: Heiko Strathmann :develop * e726ca6 / / (12 files): https://github.com/shogun-toolbox/shogun/commit/e726ca641900bf838974aafe90488bd0d0706fc111:46
shogun-notifier-shogun: Merge pull request #1017 from hushell/issues11:46
shogun-notifier-shogun:11:46
shogun-notifier-shogun: [update] Probabilistic Outputs for SVM11:46
-!- nube1 [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds]11:48
shogun-buildbotbuild #762 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/762  blamelist: hushell <hushell@hushell-U510.(none)>11:50
@iglesiasgooops11:50
@iglesiasgaah it seems nothing :)11:50
@lisitsyniglesiasg: chris is back in town (tuebingen)11:53
@iglesiasglisitsyn: how busy is he?11:55
@iglesiasgmaybe it is a good moment to ping him, no idea11:55
@lisitsyniglesiasg: he will take a look at the paper today11:56
@iglesiasglisitsyn: this is good news then!11:56
@lisitsynyeah11:56
shogun-notifier-shogun: Roman Votyakov :develop * f926f91 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/f926f91b6ae22fd37e8755b4f7301f5a8e178ca411:57
shogun-notifier-shogun: Replaced dynamic_cast with obtain_from_generic in FITCInferenceMethod class11:57
shogun-notifier-shogun: Heiko Strathmann :develop * 7c302bc / / (3 files): https://github.com/shogun-toolbox/shogun/commit/7c302bc4ae97fcce6bb1f928ddb36093666a6e1511:57
shogun-notifier-shogun: Merge pull request #1062 from votjakovr/develop11:57
shogun-notifier-shogun:11:57
shogun-notifier-shogun: Replaced dynamic_cast with obtain_from_generic in FITCInferenceMethod class11:57
@iglesiasgHeikoS has set merge mode on :)11:58
@HeikoSiglesiasg: yes :)11:58
@HeikoSits crazy11:58
@HeikoS3 days away and so many patches :)11:59
@lisitsynvotjakovr: hey, I am just curious11:59
@lisitsynare there kinda professors doing some *modern* NN research11:59
@lisitsynor do you do it by your own?11:59
shogun-notifier-shogun: van51 :develop * 0ec800c / / (3 files): https://github.com/shogun-toolbox/shogun/commit/0ec800ce08ba59659db85a450c53a3f4c1bca24d12:04
shogun-notifier-shogun: Issue #1042 - Cross-Product for lists of kernels12:04
shogun-notifier-shogun: Heiko Strathmann :develop * a641e16 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/a641e16a836b3843fae913e5416b0dd6550b050a12:04
shogun-notifier-shogun: Merge pull request #1059 from van51/develop12:04
shogun-notifier-shogun:12:04
shogun-notifier-shogun: Issue #1042 - Cross-Product for lists of kernels12:04
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has joined #shogun12:05
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/698428512:05
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has left #shogun []12:05
shogun-buildbotbuild #763 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/76312:06
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has joined #shogun12:18
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/698452112:18
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has left #shogun []12:18
shogun-buildbotbuild #1062 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1062  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, hushell <hushell@hushell-U510.(none)>12:21
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Quit: Page closed]12:26
shogun-buildbotbuild #1063 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1063  blamelist: hushell <hushell@hushell-U510.(none)>12:27
votjakovrlisitsyn: Hey :) My university's research is connected with Hierachical Temporal Memory (originally developed by Jeff Hawkins). It's a kind of biologically inspired cognitive architecture.12:29
-!- foulwall [~foulwall@2001:da8:215:503:c47e:67cf:6bd7:93a9] has quit [Remote host closed the connection]12:30
shogun-buildbotbuild #765 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/765  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, van51 <vangelis_51@hotmail.com>, Roman Votyakov <votjakovr@gmail.com>12:31
-!- travis-ci [~travis-ci@ec2-54-234-229-115.compute-1.amazonaws.com] has joined #shogun12:39
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/698468012:39
-!- travis-ci [~travis-ci@ec2-54-234-229-115.compute-1.amazonaws.com] has left #shogun []12:39
votjakovrHeikoS: Hi, There is something unclear in squared exponential covariance function with isotropic distance measure (covSEiso) in GPML matlab toolbox. There is no bug with partial derivative by ell parameter?12:44
-!- foulwall [~foulwall@2001:da8:215:202:d49d:918b:1418:6e43] has joined #shogun12:44
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun12:46
@sonney2kok I killed shogun's mailinglists12:49
@wikingboooooyaaaaa12:51
@wikingdontkiiiiliiit12:51
@sonney2ktoo late12:51
foulwall(⊙o⊙)12:52
@sonney2kI am thinking about having shogun-toolbox.org running its own mail server12:52
@wikingsonney2k: why where was it running?12:52
@sonney2kwiking, thoughts?12:53
@sonney2kmy private one12:53
@wikingsonney2k: mmm need one?12:53
@wikingsonney2k: i have an up and running mailman on my server12:53
@sonney2kneed one what?12:53
@wikingsonney2k: it's like 1 command to get a new mailinglist running on it12:53
@sonney2kwiking, I don't like mailman - hard to backup12:53
@sonney2kI mean it has a single .mbox file per mailinglist12:54
@sonney2kwhich grows and grows and grows12:54
@wikingsonney2k: well yeah i backup the whole fs :P12:54
@sonney2kso each and every day you run a backup12:54
@sonney2kyou backup *everything*12:54
@wikingsonney2k: yeah but still i can do incremental backups thanks to zfs :P12:54
@sonney2kwiking, incremental if the file is different?12:54
@wikingsonney2k: yes12:55
@sonney2khow so?12:55
@wikingsonney2k: as it's incremental on the fs level12:55
@wikingsonney2k: dont ask me, ask zfs developers12:55
@sonney2kcrazy12:55
@sonney2kmy experience with zfs was rather crash and burn12:55
@wikingi just know that i can create with one comment an incremental backup of the fs12:55
@wikingsonney2k: heheh i love it12:55
shogun-buildbotbuild #1064 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1064  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, van51 <vangelis_51@hotmail.com>, Roman Votyakov <votjakovr@gmail.com>12:55
@wikingsonney2k: but u need a lot of ram12:55
@wikingsonney2k: and it's rather unstable in linux12:56
@wikingsonney2k: hence the fbsd ;p12:56
@sonney2kahh screw it I put that thing on my server12:56
@wikingsonney2k: anyhow it's time to do something about the PRNG12:56
@wiking:D12:56
@wikingsonney2k: so shall we go with take a random element of the vector and check whether it's the same? :P12:56
@sonney2kwiking, you mean now that heiko broke all the tests you can do more ?12:56
@wikinghehehehehehe12:57
@wikingno sorry i just still populate my commints in a different branch12:57
@wiking;)12:57
@sonney2kwiking, as I said I think generating 10k numbers and checking the 10k'th to be the same with the same seed is good enough for me12:57
@lisitsynmagic number 10k ;)12:58
* sonney2k so now the question is how I get from ezmlm & qmail to mailman 12:58
@sonney2kwiking, ohh mailman has a daily patch :D12:58
@sonney2klisitsyn, does you shogun-toolbox.org mail address still work?12:59
@sonney2klisitsyn, ohh btw please send me an email under mine12:59
@lisitsyntesting12:59
@sonney2kno idea if it still works12:59
@lisitsynsonney2k: probably not12:59
@lisitsynI sent you an email12:59
@sonney2kI changed too much...13:00
@sonney2kgreylisting, spam, virtual domains, mail server blabla13:00
* sonney2k just realizes this server has ipv613:00
deerishi@sonny2k: what is sth?the "data" was not actually a directory when i git cloned shogun's repo.it was just a broken file.13:04
@wikingtada13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 4255c47 / src/shogun/mathematics/Math.h: https://github.com/shogun-toolbox/shogun/commit/4255c47b3529c4ebe407975b8841180f2de5906d13:05
shogun-notifier-shogun: Use global random as extern to avoid SG_REF/UNREF calls13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 88ce738 / src/shogun/lib/external/SFMT/SFMT.h: https://github.com/shogun-toolbox/shogun/commit/88ce7386923423202cda597aff608a6436e05c3613:05
shogun-notifier-shogun: Ignore SFMT headers in doxygen generation13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 701eb5f / src/shogun/mathematics/Random.cpp,src/shogun/mathematics/Random.h: https://github.com/shogun-toolbox/shogun/commit/701eb5f0e6e55ebaa7003e3ded3221401b9e58cb13:05
shogun-notifier-shogun: Random: use two SFMT struct instead of one13:05
shogun-notifier-shogun: SFMT random regerator requires reinitalization if one wants13:05
shogun-notifier-shogun: to generate a 64-bit random after generating a 32-bit random13:05
shogun-notifier-shogun: and vice-versa. In order to avoid reinitalization, rather have13:05
shogun-notifier-shogun: 2 different SFMT structs for the 32-bit and the 64-bit RNG.13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * c7e5f9b / src/shogun/lib/external/dSFMT/ (14 files): https://github.com/shogun-toolbox/shogun/commit/c7e5f9bf86b3824a89a351330e1f28f87335e52513:05
shogun-notifier-shogun: Initial import of dSFMT13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 0628564 / src/shogun/lib/external/dSFMT/ (3 files): https://github.com/shogun-toolbox/shogun/commit/0628564f867f78b4e5a53cebc63bd63dc65bd20d13:05
shogun-notifier-shogun: Change include paths in dSFMT for shogun13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * e659378 / src/shogun/lib/external/dSFMT/dSFMT.h: https://github.com/shogun-toolbox/shogun/commit/e659378eada060dcaaa1870c6d4738367106c02d13:05
shogun-notifier-shogun: Ignore dSFMT in doxygen generation13:05
shogun-notifier-shogun: Viktor Gal :feature/PRNG * c5f7097 / src/shogun/lib/external/dSFMT/ (3 files): https://github.com/shogun-toolbox/shogun/commit/c5f7097cdad5d1219acc9c6928a71e90cf999f6913:05
shogun-notifier-shogun: Fix redefinition of W128_T in dSFMT13:05
@wikinghehe13:05
@wikingi think shogun-notifier- went on flood protection13:05
@wiking:>13:05
@sonney2kdeerishi, well why did you git add and remove it then?13:06
@sonney2kdeerishi, it should just not be in the commit / PR13:06
deerishiyes.13:06
@wikinglet's see what travis does with this13:07
deerishihttps://github.com/deerishi/shogun/tree/m1613:07
deerishithis is the branch i sent pr from13:07
@wikingHeikoS: GMM will need a bit of redesign :P13:08
@wikingok i mean Gaussian13:08
@HeikoSwiking:  that might be13:09
@wikingHeikoS: like this for (int32_t i=0; i<m_mean.vlen; i++) random_vec[i]=CMath::randn_double();13:09
@sonney2kwiking, feel free :P13:09
@wikingHeikoS: as the new Random has functions like filling a vector of doubles13:10
@wikingfrom various intervalls13:10
@lisitsynm16 :D13:10
@wiking*intervals13:10
@lisitsynlike machine gun?13:10
@wikingbut then again u need there std distrib13:10
@wikingok i'll see what i can do about that13:10
deerishi@HeikoS: u der?13:12
@wikingno he is die :P13:12
@wikinghahahahaha this was a very lame joke :D13:12
deerishi@HeikoS: cud you look at my latest pr?13:14
@lisitsynwiking: haha nice13:15
@wikinglisitsyn: when do u arrive to berlin?13:16
@lisitsynwiking: 9th13:16
@lisitsyn21-20 or so13:16
@wikingokkey13:16
@lisitsynwiking: staying at hotel tulip13:16
shogun-notifier-shogun: Heiko Strathmann :develop * 72ba739 / src/shogun/kernel/PolyKernel.h: https://github.com/shogun-toolbox/shogun/commit/72ba73925ad80e6f98a219801170f4edcda679e513:16
shogun-notifier-shogun: added getter for degree13:16
shogun-notifier-shogun: Heiko Strathmann :develop * 9e178a5 / examples/undocumented/libshogun/ (2 files): https://github.com/shogun-toolbox/shogun/commit/9e178a577e9eae54d290ca407e458c8add50404313:16
shogun-notifier-shogun: added example how to perform modelselection over sub-parameters of combined kernels13:16
shogun-notifier-shogun: Heiko Strathmann :develop * 549256e / src/NEWS: https://github.com/shogun-toolbox/shogun/commit/549256e39861f2f735887fdc89478c90884179a913:16
shogun-notifier-shogun: mention multiclass probabilities and modelselection for combined kernels sub-parameters13:16
shogun-notifier-shogun: Heiko Strathmann :develop * 61767d9 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/61767d9d02ad5717fd24b6438ce98ceb5d0cc28513:16
@wikingi'll check my flight options13:16
shogun-notifier-shogun: Merge pull request #1066 from karlnapf/develop13:16
shogun-notifier-shogun:13:16
shogun-notifier-shogun: example for modelselection for combined kernels13:16
@lisitsynwiking: near tiergarten13:16
@wikingcool cool13:16
@sonney2ksome kind of fight going on here?13:16
@sonney2kthe one who breaks most stays until x-mas?13:16
@lisitsynhaha13:17
@iglesiasg:D13:18
@wikinglisitsyn: ok found it13:19
@lisitsynwiking: found what?13:19
shogun-buildbotbuild #766 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/766  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:20
@lisitsynder HeikoS13:22
@lisitsyn:D13:22
@HeikoSthats cyg13:23
@lisitsyndas HeikoS!13:23
@lisitsyn:D13:23
@HeikoSDER!13:23
@HeikoSdeerishi: take a look at my comments13:23
@lisitsyndie HeikoS!13:23
@lisitsyn:D13:23
@HeikoSwe just need three files from you13:23
@HeikoSsonney2k: I cannot send mails to the list anymore13:23
@lisitsynHeikoS: interesting that we have gender stuff in russian but I have absolutely no idea how to use it in german13:24
@lisitsynwell I know you are der but Kaffee?13:24
@lisitsyndas/der/die? no idea :D13:24
@lisitsynHeikoS: mailing list is der broken13:25
@HeikoSlisitsyn: der Kaffee13:25
deerishi@HeikoS:i resent the pr13:25
@HeikoSlisitsyn: broken??13:25
deerishihas sparse invariance covariance been solved?13:25
@lisitsynHeikoS: yeah der sonney2k does secret ops13:25
@lisitsynHeikoS: moving our mailing list to langley's CIA HQ13:26
@lisitsyndeerishi: yes13:26
@HeikoSdeerishi: you have to reset your git (do a backup before)13:26
@HeikoSlisitsyn:  I see13:26
@HeikoSlisitsyn: we now have modelselection for sub-kernels in MKL13:26
@HeikoSand we have probs for multiclass svm13:26
@wikingHeikoS: when u have some time13:27
@lisitsynHeikoS: I have no idea what he does  actually - check the logs for more detailed stuff13:27
@lisitsynHeikoS: yeah I noticed that for sure13:27
@wikingHeikoS: i will need your help with one example13:27
@HeikoStwo things that have been asked for often13:27
@lisitsynthe latter13:27
@HeikoSnice!13:27
@HeikoSwiking: hi!13:27
deerishibut why?13:27
@HeikoSsorry, I wasnt following13:27
@wikingstatistics_hsic.cpp13:27
@wikingHeikoS: it fails with the new prng13:27
@HeikoSdeerishi: because the PR contains things we dont want13:27
@wikingbut no idea why atm13:27
@HeikoSwiking: with the new random?13:27
shogun-notifier-shogun: Heiko Strathmann :develop * d8ededf / src/NEWS: https://github.com/shogun-toolbox/shogun/commit/d8ededf769eae4216881922ff0ad1383f40235c413:30
shogun-notifier-shogun: mention the recent streaming bugfix13:30
shogun-notifier-shogun: Heiko Strathmann :develop * 3cd0318 / src/NEWS: https://github.com/shogun-toolbox/shogun/commit/3cd0318e95f7877bd365cd115e3730892e4c30f013:30
shogun-notifier-shogun: Merge pull request #1068 from karlnapf/develop13:30
shogun-notifier-shogun:13:30
shogun-notifier-shogun: update news once more13:30
@HeikoSwiking: ok13:31
@HeikoSwiking:  remove the asserts13:31
@HeikoSI will update the example soon and move ciritcal parts to test13:31
shogun-buildbotbuild #767 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/76713:34
@sonney2klisitsyn, please rename to mr.kalashnikov13:35
@lisitsynsonney2k: why that?13:35
@sonney2kyou are a killer13:36
deerishi@HeikoS:i resent the pr13:36
@lisitsynsonney2k: people around won't understand why :D13:36
@HeikoSdeerishi: there are still the data commits in there13:37
@HeikoSyou have to reset your git version13:38
@HeikoSotherwise they wont go away13:38
@HeikoSbut backup your code before13:38
@lisitsynsonney2k: have you ever disassemble and assemble AK-47 btw? :D13:39
@lisitsynas you lived in not the best part of berlin you may have that too13:39
@sonney2ksonney2k, I do that blindfolded lefthanded in 10 seconds!13:39
@sonney2kyou learn that when you protect the wall(tm)13:40
@lisitsynsonney2k: my mother did that in a minute or so :D they have to do that in school13:40
@lisitsynooh wait I did that too13:40
@sonney2kso well earned13:40
@sonney2kdeserved13:40
@lisitsynsonney2k: where the wall was?13:41
@lisitsynwhich street?13:41
@HeikoShaha, you should use your skills to guard our workshop sergey :)13:41
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has joined #shogun13:41
travis-ci[travis-ci] it's Viktor Gal'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/698587613:41
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has left #shogun []13:41
@HeikoSwiking: does removing the asserts help?13:41
@lisitsynHeikoS: I did that but I am terribly bad in it I think :D13:41
@lisitsynHeikoS: is friedrichstrasse the street the wall was at?13:43
votjakovrHeikoS: Hi, I'd like to ask: is there any bug in covSEiso (GPML matlab toolbox)?13:51
shogun-buildbotbuild #1065 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1065  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:52
@sonney2klisitsyn, yes that too13:53
shogun-buildbotbuild #1066 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1066  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:56
-!- van51 [~van51@athedsl-318776.home.otenet.gr] has joined #shogun14:01
shogun-buildbotbuild #769 of cyg1 - libshogun is complete: Failure [failed configure]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/769  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:02
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Quit: Page closed]14:11
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun14:11
deerishi@HeikoS: i am doing git reset14:14
deerishigit reset --soft .but what changes should i see in my repo14:14
@sonney2kman if you want google to help you with mailman setup you find that http://www.dailyrecord.co.uk/news/crime/mailman-exposed-himself-motorists-mowed-171657514:14
@iglesiasghahaha14:15
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Quit: Page closed]14:16
@sonney2kHeikoS, btw did you get some kind of email that you couldn't post to the list or so?14:17
-!- nube [~rho@49.244.99.69] has joined #shogun14:21
shogun-buildbotbuild #1067 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1067  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>14:23
-!- nube [~rho@49.244.99.69] has quit [Quit: Leaving.]14:33
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds]14:37
@wikingHeikoS: well if i remove the assert of course it'll fix it :P14:37
@HeikoSsonney2k: yes two14:40
@HeikoSwiking: I will update the stuff soon, could you add a github issue and assign it to me?14:40
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has joined #shogun14:43
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/698646114:43
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has left #shogun []14:43
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun14:46
@wikingHeikoS: ah but i thought i can fix it :)14:47
@HeikoSwiking: well its a stupid test14:47
@HeikoSso only way to fix it is to rewrite :)14:47
@wikingah no shit14:47
@HeikoSfix data and assert correct numbers14:48
@wikingwhy is it there in the first place!? :)14:48
@HeikoSsince I was not thinking properly14:48
@wikingclearly14:48
@wiking:)14:48
@HeikoS:)14:48
@wikingCLEARLY14:48
@wikingok i have to remove it because i want to see what travis does with the unit tests with the new PRNG14:49
@wikingbut it's only from feature/PRNG branch14:49
@sonney2kHeikoS, please forward to me14:52
@HeikoSsonney2k:  dont you get a copy anyway?14:52
deerishi@HeikoS : i did git reset --hard.then i did git commit -m "kpca".it said nothing to commit.working directory clean14:53
shogun-notifier-shogun: Viktor Gal :feature/PRNG * e270af7 / examples/undocumented/libshogun/Makefile: https://github.com/shogun-toolbox/shogun/commit/e270af7eff0a9a9c6333e4861001eee375fda94914:53
shogun-notifier-shogun: Remove statistics_hsic example run14:53
@wikingmmm let's see what travis does now14:53
@HeikoSsonney2k: could you help deerishi, no idea whats the problem there14:54
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Ping timeout: 245 seconds]14:58
@sonney2kdeerishi, --hard erases15:02
@sonney2kyou need to do --soft15:02
deerishiwhat changes do i expect to see after reset in the repo15:03
@sonney2kdeerishi, all of them15:03
@wikingok lol15:04
@wikingi have not seen this before15:04
@wikinghttps://travis-ci.org/shogun-toolbox/shogun/jobs/698838715:04
@sonney2kwiking, there is too much to wonder about15:05
deerishi@sonny2k: i did git reset --soft in my main repo.after that i did git commit -m ,then git push origin branchname15:05
deerishiit still came,everything is upto date15:06
@sonney2kdeerishi, use git push --force origin15:08
deerishi@sonny2k :the data commits still seem to there15:12
@sonney2kdeerishi, well if you did commit everything ofcourse15:13
@sonney2kdeerishi, you should do git reset --soft <id>15:13
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has joined #shogun15:13
@sonney2kthen remove unnecessary files15:13
@sonney2kor git commit only the needed files15:14
@lisitsynsonney2k: git consulting company sonnenburg and partners15:15
@lisitsyn;)15:15
@sonney2kwiking, I *hate* exim & mailman to the deepest of my heart now15:16
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has joined #shogun15:16
travis-ci[travis-ci] it's Viktor Gal'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/698838515:16
-!- travis-ci [~travis-ci@ec2-107-21-134-219.compute-1.amazonaws.com] has left #shogun []15:16
@sonney2kman is qmail *simple* *easy*15:16
@sonney2k*fast*15:16
@lisitsynsonney2k: why did you go away from simple easy fast thing - I didn't get the point15:16
@sonney2ktoo bad the author is an %@^%$#!15:16
@sonney2klisitsyn, last update of qmail is 10 years ago or so15:17
@lisitsynhaha15:17
@sonney2k0 maintainance15:17
@lisitsynstable release 199815:17
@lisitsynlol!15:17
@sonney2krunning on a machine with 128MB15:17
@wikingsonney2k: they are great ;)15:17
@wikingsonney2k: u just need to tweak them :DDDD15:17
vgorbatilisitsyn: hi, sorry for being afk that long - holidays + preparation for the bachelor's work took too much time:)15:19
@wikingHeikoS: any ideas why this could happen: statistics/QuadraticTimMD_unittest.cc:188: Failure15:19
@wikingExpected: (CMath::abs(mean-CStatistics::mean(null_samples))) <= (10E-8), actual: 0.043256 vs 1e-0715:19
@wikingstatistics/QuadraticTimMD_unittest.cc:189: Failure15:19
@wikingExpected: (CMath::abs(var-CStatistics::variance(null_samples))) <= (10E-8), actual: 0.0285305 vs 1e-0715:19
@sonney2kwiking, HeikoS domain...15:20
@lisitsynvgorbati: I see - just let me know when you are not going to be around15:20
@sonney2kI guess rng sucks :D15:20
@wikingahhahahah15:20
@sonney2kvgorbati, before you are gone :D15:20
vgorbatilisitsyn: I thought that you should have these holidays too:)15:21
@wikingah ok i think i know now15:21
@wikingalthough again a funny feature that it doesn't fail on my machine15:21
@sonney2kwiking, ?15:21
@wikingsonney2k: he reinits the PRNG15:22
@lisitsynvgorbati: not only me but other guys :)15:22
@sonney2kvgorbati, we have no holidays15:22
@sonney2kwe don't even sleep15:22
@lisitsynhah15:22
@lisitsynbut actually I was here on holidays15:22
@wikingCMath::init_random(1);15:22
@wikingcalls twice15:22
-!- abinayam [3d0c137b@gateway/web/freenode/ip.61.12.19.123] has joined #shogun15:22
@wikingso i'll have to change a little bit the code :P15:22
vgorbatisonney2k: robots, or something?)15:22
@lisitsynsonney2k: I am going to have holidays tomorrow due to our old deal you know15:23
@lisitsyn:D15:23
@sonney2kvgorbati, exactly15:24
@wikingyeeey i've managed to break it!15:24
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has joined #shogun15:24
-!- foulwall [~foulwall@2001:da8:215:202:d49d:918b:1418:6e43] has quit [Ping timeout: 245 seconds]15:25
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]15:31
Yanglittleexcuse me , i installed the shogun, but it says can't find the module , does the version of python must be 2.5 or 2.6?15:31
@sonney2kYanglittle, anything from 2.4 should work15:32
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has joined #shogun15:32
travis-ci[travis-ci] it's Viktor Gal'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/698839515:32
-!- travis-ci [~travis-ci@ec2-23-20-196-170.compute-1.amazonaws.com] has left #shogun []15:32
@sonney2kwiking, congrats!15:32
@sonney2kwiking, people with your talent are rare these days :P15:33
Yanglittle:)15:45
@wikingsonney2k: breakers? :)15:49
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds]15:52
erlendaexit15:53
-!- erlenda [~erlenda@fw-oslo.intelcom.no] has quit [Quit: Leaving]15:53
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun15:59
-!- mode/#shogun [+o iglesiasg] by ChanServ16:00
@iglesiasgsonney2k: ping16:10
Yanglittleit still can't impot the module, and says no the module16:10
@iglesiasgYanglittle: just with MulticlassLabels or also with others?16:12
Yanglittle all ,  shogun.Features, shogun.Classifier16:13
Yanglittleall can't find the module16:13
@iglesiasgYanglittle: so you are not able to execute any of the examples in examples/undocumented/python_modular I guess?16:14
@iglesiasgYanglittle: have you compiled, built and installed shogun with python_modular support?16:14
@iglesiasgsth like: ./configure --interfaces=python_modular16:14
@iglesiasgmake16:14
@iglesiasgmake install16:15
Yanglittlei run ./configure  make  make install16:16
@iglesiasgYanglittle: you need to specify python as I have written above16:17
Yanglittleok.16:18
@iglesiasgYanglittle: otherwise shogun is not built with python support16:18
@iglesiasgYanglittle: note also that install might require special permissions (i.e. sudo make install)16:18
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun16:25
@HeikoSwiking16:26
@HeikoSthe broken travis is again due to stupid unit tests by me16:27
@sonney2kHeikoS, could you please send your email to the list again?16:29
@sonney2kiglesiasg, ?16:29
deerishi@HeikoS: i just sent another PR.hopefully all the data commits are missing16:29
@iglesiasgsonney2k: just wanted to let you know that I might be a bit absent until Saturday, when I doing my (hopefully) last exam towards my degree :)16:30
@sonney2kdeerishi, look it is again 9 commits16:30
@sonney2kthen good luck!16:30
@HeikoSiglesiasg: good luck with that16:31
@iglesiasgsonney2k: I am planning to do a contribution soon anyway, something for the hmsvm16:31
@HeikoSdeerishi: you can see yourself whether there is something wrong before sending the PR16:31
@iglesiasgsonney2k, HeikoS : thanks!16:31
@sonney2kHeikoS, please send the mail again!16:31
@iglesiasgit is about automata theory, somewhat entertaining16:31
@HeikoSsonney2k: I just did16:32
@sonney2kHeikoS, please now!16:32
* sonney2k checks16:32
@HeikoSdidnt come back16:32
@sonney2kvery good16:32
@sonney2kgreylisted16:32
@sonney2kso it might arrive in >10mins16:32
@HeikoSok?16:33
@HeikoSwiking, I correct myself, something is strange with the broken thing16:33
@sonney2kI thHeyes16:33
@sonney2kHeikoS, yes16:33
@HeikoSwiking16:34
@HeikoSthere is CMath::init_random(1); in line 16116:34
@HeikoSif you also put that line into 175 it might work16:35
@sonney2kHeikoS, did you get my email?16:42
@HeikoSsonney2k: yes16:42
@sonney2kk16:42
-!- abinayam [3d0c137b@gateway/web/freenode/ip.61.12.19.123] has quit [Ping timeout: 245 seconds]16:45
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Quit: Page closed]16:49
-!- gsomix [~Miranda@188.168.3.6] has joined #shogun16:51
gsomixhi, guys16:51
-!- van51 [~van51@athedsl-318776.home.otenet.gr] has quit [Quit: Leaving.]16:53
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has quit [Ping timeout: 245 seconds]17:10
Yanglittlei installed python 2.4 and numpy, but it can't find the numpy when running ./configure...17:16
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has quit [Quit: Page closed]17:28
-!- nube [~rho@49.244.120.94] has joined #shogun17:28
@wikingHeikoS: yeah i saw that17:30
@wikingHeikoS: you mean line 182 or?17:31
@wikingHeikoS: bootstrap_null uses CMath::random in some way i guess17:32
@wiking?17:32
-!- foulwall [~foulwall@2001:da8:215:c252:d15d:f218:cac9:393e] has joined #shogun17:35
-!- nube [~rho@49.244.120.94] has quit [Quit: Leaving.]17:48
-!- deerishi_ [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun17:51
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has joined #shogun17:52
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds]17:52
pickle27sonney2k: so the merge is all good then?17:53
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]17:53
deerishi_@sonny2k : u told me to send the pr as a single commit17:57
deerishi_was that the only problme in the pr?or were there any data commits too?17:58
deerishi_HeikoS:  were there any data commits  in the pr?18:08
deerishi_sonny2k told me to send them as a single git commit18:08
deerishi_@HeikoS : u der?18:11
@wikingdeerishi_: trust me people will answer u when they are here18:11
@wikingotherwise they wont18:11
deerishi_k18:11
@wikingneed to be more patient18:11
deerishi_allright.i was just wanted to get over this pr.the code is working fine,but m having some issues with the merging part18:12
deerishi_i'l b more patient18:13
deerishi_:)18:13
deerishi_@wiking : do you have any idea on how to send as a single commit18:18
deerishi_if i have modified 4 files?18:19
@wikingwell git commit -a18:19
@wikingthat'll send all the modified files into the commit18:20
deerishi_then i should simply do git push origin branchname,rite?18:27
@wikingy18:27
-!- deerishi_ [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Quit: Page closed]18:44
-!- gsomix [~Miranda@188.168.3.6] has quit [Read error: Connection reset by peer]18:45
@sonney2kHeikoS, hurray! so mailinglist works again \o/18:48
-!- gsomix [~gsomix@188.168.3.6] has joined #shogun18:49
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has joined #shogun18:51
deerishi@wiking : i sent the pr again,but it still showed 11 commits instead of 118:52
-!- foulwall [~foulwall@2001:da8:215:c252:d15d:f218:cac9:393e] has quit [Remote host closed the connection]19:00
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun19:03
shogun-notifier-shogun: van51 :develop * 9e2afda / / (4 files): https://github.com/shogun-toolbox/shogun/commit/9e2afdad9387da2e7f13e8771bbbe4129a906c0319:03
shogun-notifier-shogun: Update on #1042. Fixed order of returned kernels19:03
shogun-notifier-shogun: Heiko Strathmann :develop * 2a8b6f1 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/2a8b6f124a51e847832eb02566cfcea35af6617719:03
shogun-notifier-shogun: Merge pull request #1070 from van51/develop19:03
shogun-notifier-shogun:19:03
shogun-notifier-shogun: Update on #1042. Fixed order of returned kernels19:03
@HeikoSdeerishi: never do git commit -a19:04
@HeikoScommit all files by hand19:04
@HeikoSotherwise you will always commit these data things19:04
deerishiThe PR i had sent ,did it contain any data commits?i think i had removed all the data commits before sending the pr.19:05
deerishiall the broken files named data,i had removed19:05
@HeikoSdeerishi: you can see this yourself19:06
@HeikoSjust click on diff19:06
deerishialso i need your advice ,on the next bug .as in what should do next?19:08
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 43af113 / src/shogun/mathematics/Math.cpp,src/shogun/mathematics/Math.h: https://github.com/shogun-toolbox/shogun/commit/43af113771b39b83fd4409f40ef28f9bf67bab2719:09
shogun-notifier-shogun: Change CMath::init_random to use CRandom's set_seed19:09
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 9ea5c08 / tests/unit/statistics/QuadraticTimeMMD_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/9ea5c08da92a1a9d6f0492676c9e3ad77aa2e27f19:09
shogun-notifier-shogun: Fix QuadraticTimeMMD unit test19:09
shogun-notifier-shogun: the seed wasn't set in the right way19:09
deerishi@HeikoS: also sonny2k told me to submit as a single commit19:10
@wikingletssee if that one fixed it19:10
@HeikoSdeerishi: yes that is all three files, not more, so add them by hand with one single commit, dont do commit -a19:10
@HeikoSdeerishi: what do you mean by "on the next bug?"19:11
deerishii want to contribute more.should i start working on the proposal or should i solve more issues19:11
@HeikoSdeerishi: so the first thing you should do is to get this patch submitted19:17
@HeikoSsince proposal deadline is over, the only thing that is left is to submit more good patches19:18
shogun-buildbotbuild #770 of cyg1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/77019:20
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun19:24
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/699534719:24
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun []19:24
deerishi@HeikoS: this time i manually added the files: git add -f [names of the 4 files] ,them did git commit -m "kpca",then git push origin branchname19:37
shogun-buildbotbuild #1068 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1068  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:38
deerishi@HeikoS : could you take a look at this https://github.com/deerishi/shogun/compare/mm519:38
-!- lisitsyn [~blackburn@188-122-243-31.clients.tlt.100megabit.ru] has joined #shogun19:41
-!- mode/#shogun [+o lisitsyn] by ChanServ19:41
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun19:41
travis-ci[travis-ci] it's Viktor Gal'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/699554619:41
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun []19:41
@wikingmmmm can i serialize actually mutliclassliblinear?19:47
@sonney2kwiking, you will find out when you tried :D19:49
@wikingsonney2k: i have like a fucking 300M file19:50
@wikingserialized out19:50
@sonney2kwiking, ohh yeah19:50
@sonney2kthere is room for improvement19:50
@sonney2kwe save everything multiple times19:51
@sonney2kif e.g. you have 10 different kernels all using the same features19:51
@sonney2k-> 10times the features19:51
@sonney2kdeerishi, one suggestion:19:51
@sonney2kdeerishi, copy the files you intended to submit19:51
@sonney2kthen create a fresh clone19:51
@sonney2kcopy the files back into the clone19:51
@sonney2kdo a git commit and push19:52
@sonney2kand do a PR19:52
@lisitsynwiking: if you had 2x more classes19:52
@lisitsynyou would have 600M19:52
@lisitsyn:D19:52
deerishi@sonny2k : but wont that also get back the data commits too.19:52
@sonney2kHeikoS, just to get some feedback19:52
deerishi@sonny2k : could you take a look at this https://github.com/deerishi/shogun/compare/mm519:52
@sonney2kdeerishi, no19:52
@sonney2kdeerishi, yeah it is horrible just don't touch all these data etc files19:53
@sonney2kHeikoS, say we wanted to improve serialization by avoiding duplicates19:53
@sonney2klisitsyn, and that yes19:53
@sonney2kHeikoS, so how would we do it - since you are the master of disaster - I would say we introduce some running number of objects19:54
shogun-buildbotbuild #1069 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1069  blamelist: van51 <vangelis_51@hotmail.com>19:54
@sonney2kwhenever objects are the same (meaning same pointer!) they will just be referenced by the running number ptr19:55
@sonney2klisitsyn, have you seen? the mailinglist works again :D19:55
@wikinglisitsyn: when i deserialize the machine the w (one of the class) equal original feature size + numclasses19:55
@sonney2kohh and new record19:55
@sonney2k315 subscribers19:55
@lisitsynsonney2k: I received test message19:55
@sonney2kyeah and I the answer19:56
@lisitsynwiking: oops?19:56
@sonney2kso I think mail works fine again19:56
@wikinglisitsyn: yeah i have no clue why is that happening19:56
@wikinglisitsyn: but i'm getting an assertation fail19:56
@lisitsyndissertation fail19:56
@wikingthat w.vlen != feautresize19:56
@lisitsynwe need a hero!19:57
@lisitsyn:D19:57
@sonney2kHeikoS, does that sound reasonable? back then when you wrote the migration framework you did have to load all original data anyways19:57
@sonney2kHeikoS, so the same thing would have to be done and then the objects need to be filled19:58
@sonney2kwiking, let me guess - sparse data?19:58
@sonney2kwiking, btw one way to make the model small is the clear the features away19:59
@sonney2kwiking, then you just have the w19:59
@sonney2k's19:59
@wikingsonney2k: noup dense19:59
@wikingi mean the very interesting thing is that multiclassliblibnear takes like 5+ minutes to train 20k vectors with 170 features20:01
@lisitsynwiking: could be faster..20:05
@lisitsynwiking: how many classes?20:05
@HeikoSsonney2k: I gotta go now, sorry, could you send me an email with this? didnt get it from the chat...20:06
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun20:06
-!- van51 [~van51@77.49.220.109.dsl.dyn.forthnet.gr] has joined #shogun20:06
deerishi@sonny2k: @HeikoS  told me to remove all the data commits,if i say what you suggested then the data commits will still reamin20:07
deerishi*remain20:07
@sonney2kHeikoS, it was just about avoiding duplicates20:07
@sonney2kpickle27, did you already refactor that train function into some smaller sub-units?20:08
@sonney2kit's been a bit messy20:08
@sonney2kdeerishi, look current git develop is OK - so if you just git add your 3 files to that and commit & push you have a perfectly legit PR20:09
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 43028bf / examples/undocumented/libshogun/evaluation_cross_validation_multiclass_mkl.cpp: https://github.com/shogun-toolbox/shogun/commit/43028bfaa4d9cd3baf1906080210fcdc4c782c0320:09
shogun-notifier-shogun: Fix evaluation_cross_validation_multiclass_mkl.cpp for new PRNG20:09
@sonney2kso just stop wondering and do it and be happy!20:09
deerishialright.i will do that again.although i have done that a couple of times before.20:13
-!- viksvampire19_ [~chatzilla@115.184.13.194] has joined #shogun20:14
-!- viksvampire19_ is now known as viksvampire1920:14
viksvampire19hi sonney2k and lisitsyn20:17
@lisitsynhey20:19
viksvampire19As I have submitted my proposal for GSOC , I need ur reviews on it n what should I do before the results come out?20:21
gsomixlisitsyn, yo, maaaan20:21
@lisitsynviksvampire19: feel free to contibute to support your proposal with real code20:22
viksvampire19ok..so I should start coding on the same .. I mean I have coded the general csv reader that supports SGMatrix..I will put it up at my github profile..therz only a simple ascii reader..20:25
@lisitsynplease issue a PR so we can comment and discuss it20:25
viksvampire19sure..I gues I have issued PR..20:26
-!- van511 [~van51@athedsl-408474.home.otenet.gr] has joined #shogun20:27
-!- viksvampire19 [~chatzilla@115.184.13.194] has quit [Quit: ChatZilla 0.9.90 [Firefox 20.0/20130329030832]]20:29
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has joined #shogun20:30
travis-ci[travis-ci] it's Viktor Gal'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/699738120:30
-!- travis-ci [~travis-ci@ec2-107-22-117-126.compute-1.amazonaws.com] has left #shogun []20:30
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds]20:30
-!- deerishi [73f88294@gateway/web/freenode/ip.115.248.130.148] has quit [Ping timeout: 245 seconds]20:30
-!- van51 [~van51@77.49.220.109.dsl.dyn.forthnet.gr] has quit [Ping timeout: 276 seconds]20:31
-!- van511 [~van51@athedsl-408474.home.otenet.gr] has quit [Client Quit]20:31
-!- van51 [~van51@athedsl-408474.home.otenet.gr] has joined #shogun20:32
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has quit [Ping timeout: 245 seconds]20:33
-!- van51 [~van51@athedsl-408474.home.otenet.gr] has quit [Ping timeout: 268 seconds]20:45
-!- iglesiasg [d58f32e4@gateway/web/freenode/ip.213.143.50.228] has joined #shogun21:06
@wikingmmmm lisitsyn ^ see that...21:06
@wikinglisitsyn: https://travis-ci.org/shogun-toolbox/shogun/jobs/699738321:06
-!- mode/#shogun [+o iglesiasg] by ChanServ21:06
@lisitsynwiking: can't see what failed21:07
@wikinglisitsyn: ff?21:07
@wikinglisitsyn: IsomapTest.neighbors_preserving21:07
@lisitsynohhh21:07
@wikingconverter/Isomap_unittest.cc:194: Failure21:08
@wikingExpected: ((float64_t) similar_elements_count /(float64_t) total_elements_count) >= (min_similarity_level), actual: 0.4 vs 0.821:08
@wiking#similarElements/#total < minimal similarity level.21:08
@wikingi mean it's with the new PRNG21:08
@wikingand funny enouhg21:08
@wikingit only fails with clang on travis21:08
@wikinggcc went w/o problems :D21:08
@wikingand of course on my machine it works w/o problems :D21:09
@lisitsynvgorbati: хеллоу21:10
vgorbatilisitsyn: hi)21:11
@lisitsynvgorbati: houston has got a problem21:11
vgorbatilisitsyn: what problem?21:11
@lisitsynvgorbati: check previous messages21:11
@lisitsyn:)21:11
@lisitsynisomap tests fails sometimes now21:12
@lisitsyn:D21:12
@lisitsynwiking: okay it is not landmark isomap21:12
@lisitsynso it is not random based21:12
vgorbatilisitsyn: it can't be that bad21:13
vgorbatilisitsyn: I mean 0.421:14
@lisitsynvgorbati: yeah I wonder what happened21:14
@lisitsynwiking: so ONLY clang + travis + prng21:14
@lisitsynwhat is so special in travis? why it fails big time21:15
-!- iglesiasg [d58f32e4@gateway/web/freenode/ip.213.143.50.228] has quit [Quit: Page closed]21:23
vgorbatilisitsyn: I do not even what to say:) There is even 0.1 similarity level - it is 9 of 10 neighbors missing:) From your previous discussion I see that it can be because of new random number generator, so maybe switch the test to be fully deterministic?21:24
@lisitsynvgorbati: would be nice if we could reproduce the error21:25
@lisitsynvgorbati: is there anything you are working on (shogun/tapkee wise)?21:25
vgorbatilisitsyn: well, I am kind a working on the manifold sculpting, but I guess I can put it on hold21:26
vgorbatilisitsyn: but I have no idea how to reproduce this issue)21:27
@lisitsynvgorbati: ah! remember :D21:27
@lisitsynhow is it going?21:27
vgorbatilisitsyn: Given that there are 5 algorithm steps (more or less comparable), I am approaching the 3rd one21:28
@lisitsynI see21:29
@wikinglisitsyn: no idea21:41
@wiking:(21:41
@lisitsynwiking: okay I'll take a look again tomorrow21:43
@lisitsynif you can just ignore it for now21:44
@wikinglisitsyn: yeah i mean i easily ignore it21:46
@wikingjust that i dont know why it's happening21:46
@wikingbut yeah the PRNG branch is growing big time21:46
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun21:46
@wikingit'll need some fair amount of merging :P21:46
shogun-notifier-shogun: Viktor Gal :feature/PRNG * 01c0bea / examples/undocumented/libshogun/Makefile: https://github.com/shogun-toolbox/shogun/commit/01c0bea0b8e5bffd435a8ebec76f2997b8bce7eb21:48
shogun-notifier-shogun: Add back temporarily removed statistics_hsic21:48
@wikinglisitsyn: lol21:55
@wikinghttps://travis-ci.org/shogun-toolbox/shogun/jobs/700022121:55
@wikingit worked fine now21:55
@lisitsynwiking: ohh21:55
@lisitsyna chance!21:55
@lisitsynwiking: you won a million!21:55
@wikingok i'll have to run like kazillion times21:56
@lisitsynwiking: hmm may be graph was disconnected21:56
@wikingbuildbot on this21:56
@lisitsynwe can fix that I think21:56
@wikingok there's now the integration tests failing21:57
@wikingbut that's kind of like21:57
@wikingnormal21:57
@wikingas we saved the data for another PRNG21:57
@lisitsynyeah21:57
@wikinggcc failed21:57
@wikinghalleluya21:57
@wikingfuckthisshit21:58
@sonney2kwiking, haha21:58
@wikinghahaha21:58
@wikingi know why21:59
@wiking:D21:59
@sonney2kenlighten us :)21:59
@wikingat least!21:59
@wikingok so the order of unit tests are different21:59
@sonney2kahh ...21:59
@wikingand there's for sure some random somewhere21:59
@wikingbut still i think21:59
@wikingit should never fail21:59
@wikingregardless of the PRNG state22:00
@wikingbut yeah i mean IsomapTest.neighbors_preserving uses PRNG22:00
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving]22:00
@wikingand in the case where it doesn't fail22:01
@wikingprng is reseeded22:01
@wikingat least twice22:01
@wikingwhile in the failed case it's not22:02
@wikingok i think this is some prng error22:02
@wikingit has to be22:02
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Quit: Page closed]22:02
@wikinglisitsyn: if the test features in IsomapTest.neighbors_preserving are from std distrib22:02
@wikingit should never fail22:02
@wikingright?22:02
@lisitsynwiking: I can't guarantee that :D22:03
@wikingmmmm22:03
@wikinglisitsyn: should i fix the seed then there?22:03
@lisitsynwiking: yeah something has to be done here22:03
@wikingfill_matrix_with_test_data22:03
@lisitsynwiking: yeah if that helps22:03
@wikingbecause u use CMath::randn_double22:04
@lisitsynwiking: later I think I have an idea what to do22:04
@wikingok i'll try to fix the seed there22:04
@wikingand i'll try different seed22:04
@wikingmaaaybe i find a by chance a seed that makes it fail22:04
@lisitsynwiking: the thing is that it constructs neighborhood graph22:04
@lisitsynand if it is disconnected22:04
@lisitsynthe result won't be much helpful22:05
@lisitsynsome nans may be22:05
@lisitsynI either could throw an exception if it is not connected22:05
@wikingmaaaybe i find a by chance a seed that makes it fail22:05
@wikingmmmm22:05
@wikingno not here :)22:05
@sonney2klisitsyn, well you should seed with something that works then22:05
@wikingi'm just trying different seeds22:05
@sonney2kand then it is ok22:05
@lisitsynsonney2k: I think I need a routine that connects components22:06
@lisitsynor selects the largest (in means of size) component22:06
@wikingok i'm trying now the 5th differnt seed22:06
@wikingcouldnt find a failing one22:06
@wikinghow can we fix the order of unit tests?22:06
@wikingmaybe i can then reproduce the thing22:06
@sonney2kwiking, IMHO before each unit test is run a seed should be set22:07
@sonney2kor in each unit test that has some randomness22:07
@wikingi mean it's a good thing to have some randomness imho22:08
@wikingas this case22:08
@wikingbecause maaaybe i'm still doing something wrong22:08
@wikingin the PRNG22:08
@wikingthe comment for SFMT and dSFMT is done by japanese22:08
@wikingand their english is not so much clear in some cases22:09
@wiking:D22:09
@wikingsonney2k: btw for these apis22:10
@wikingrandom(float32_t min_value, float32_t max_value)22:10
@wikingcan we do that it's (min_value, max_value) and not the closed interval22:10
@wikingas that we the PRNG is much faster22:10
-!- travis-ci [~travis-ci@ec2-184-72-78-247.compute-1.amazonaws.com] has joined #shogun22:10
travis-ci[travis-ci] it's Viktor Gal'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/700021922:10
-!- travis-ci [~travis-ci@ec2-184-72-78-247.compute-1.amazonaws.com] has left #shogun []22:10
@wikingi mean that the actual return value of that function is in (min_value, max_value) and not [min_value, max_value]22:11
vgorbatilisitsyn: here?22:19
@lisitsynvgorbati: yes22:19
vgorbatilisitsyn: I wanted to ask: in the algorithm Gashler says, that he preprocesses the data using PCA, but without reducing the dimensionality: he rotates the points around first n_target_dim principal components.22:22
@wikingsonney2k: 0/1? :)22:22
vgorbatilisitsyn: I am not sure what he means by this. As I see it, you can't rotate around 2 vectors in 3-D space, you need 3 vectors22:22
vgorbatilisitsyn: And since PCA gives perpendicular eigen-vectors, the only way that I see to rotate while preserving dimensions is to use all eigen-vectors to project the data22:23
@wikingwho fscked up regression? :)22:24
@lisitsynvgorbati: yeah makes sense to me22:24
@lisitsynvgorbati: I do not actually realize how to use only target_dim components here22:26
@lisitsynit won't be preprocessing as it already reduces everything22:26
@wikingwhere is votjakovr22:26
@lisitsynwiking: barnaul is asleep I think22:27
@wikingsonney2k: new measure for deciding who gets that 8 slots: number of times you break travis :)22:27
@lisitsynwiking: just check where it is :D22:27
@wikingthe top 2 guys gets out from the game22:27
@lisitsynnot china but half a way22:27
@wikingvotjakovr++22:27
@wiking:)22:27
vgorbatilisitsyn: huuh, because I actually spend quite a lot of time trying to understand how he does it, and did not get it at all:) Maybe an algorithm that he gives looks familiar for you: it is here http://axon.cs.byu.edu/papers/gashler2011smc.pdf, on the second last page22:28
@wikingbut of course if you fix somebody else's error you get -122:28
@wiking;)22:28
@lisitsynvgorbati: let me check22:29
@wikinghence u can correct your earlier mistakes :)22:29
@wikingsonney2k: ^ totally good measure :P22:29
@wikinglisitsyn: i'll hack your github accout and break the whole shogun :))022:29
@lisitsynwiking: uhmm and?22:29
@lisitsyn:D22:29
@wikingyou'll have like +1000 :D22:30
@lisitsynwiking: aaand?22:30
@wikingand you are out22:30
@wiking:)22:30
@lisitsynwiking: out of?22:30
@wikinggsoc22:30
@lisitsynwiking: so I won't be admin anymore?22:31
* wiking is tired...22:31
@lisitsyn:D22:31
@wikinglisitsyn: nada22:31
@wikingzilch22:31
@wikingwe'll have to find another russian slave to do all the work22:31
@wiking:S22:31
gsomixhttp://www.youtube.com/watch?v=g3qkbc28R_I22:31
@wikinggsomix: wtf is this? :D22:32
gsomixit's fine22:32
@wikingallzgood22:32
@wikingdamnshiats22:32
@wikingwe'll never have trees in shogun :(22:32
@wikingi saw this and i got facepalmed22:32
@wiking#32522:33
@wikingDo you expect a code translation from java from weka to C++?22:33
@lisitsynnot sure what she means22:33
@lisitsynvgorbati: again , page 2?22:34
@wikinglisitsyn: well i guess it's like 'manual translation'22:34
vgorbatilisitsyn: second from the end22:34
@wikinglisitsyn: hope not some codegenerator :))))22:34
@lisitsynwiking: it is pain in da azz22:34
@wikinglisitsyn: hehehehe22:34
@wikinglisitsyn: that'd be awesomeness :D22:34
@wikinglets start codegen evertyhing into shogun22:34
@wiking:DDD22:34
@lisitsynvgorbati: do 20 times?22:35
@lisitsyn:D22:35
@lisitsynoh I recognize gashler22:35
vgorbatilisitsyn: as I said, he loves magic constants)22:35
@wikinglisitsyn: like /me wantz linux kernel in shogun22:35
gsomixwiking, codegenerator of codegenerators22:35
@wikinggsomix: yeah !!!!22:35
@wikingand swig it!22:35
@wikingas always22:35
@lisitsynvgorbati: did I told you he computes Ax 100 times to compute largest eigenvector?22:35
@wiking:)22:35
@lisitsyn100? what the hell..22:36
@lisitsyndo 20 times22:36
@lisitsynargh22:36
@lisitsynvgorbati: I do not recognize the algorithm22:37
vgorbatilisitsyn: :D This algorihtm just makes very little sense to me)22:37
vgorbatilisitsyn: if any22:37
@lisitsynit is like gram-schmidt stuff22:37
@lisitsyn20 times22:37
@lisitsyn:D22:37
@lisitsynvgorbati: as I understand22:38
@lisitsynit is kind of optional22:38
@lisitsynjust to help that thing to converge22:38
@lisitsynright?22:38
vgorbatilisitsyn: yes22:38
@lisitsynvgorbati: I see something from GS orthogonalization actually here22:39
vgorbatilisitsyn: what does gramm-schmidt process do?22:39
@lisitsynvgorbati: finds an orthogonal basis from a set of vectors22:39
@lisitsynbasically the same as QR decomposition22:40
vgorbatilisitsyn: but he uses it on principal components - what is the need, or do I miss something?22:40
@lisitsynvgorbati: no I don't mean it does the same - I just mean it resembles some parts of it22:41
@lisitsynwell any iterative projection would22:41
vgorbatilisitsyn: you are right, because in the code there's even a comment: "Use the modified Gram-Schmidt process to compute the other plane vector")22:42
@lisitsynhah lucky guess22:42
@lisitsynvgorbati: I have no idea how it was published :D22:43
@lisitsynmy experience of JMLR submission tells reviewers would not like that22:44
@lisitsynwhy 20 times :D22:44
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has joined #shogun22:45
@lisitsynhow can that be fixed in pseudocode..22:45
vgorbatilisitsyn: :) I guess I will skip this step, or maybe just rotate as we discussed?22:45
@lisitsynvgorbati: just skip it for now22:45
@lisitsynvgorbati: actually that 20 times step reminds me power iteration method22:45
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has joined #shogun22:46
@lisitsynvgorbati: okay just my thoughts on that algorithm22:47
@lisitsynwe subtract mean from the data to center it (it is standard PCA stuff)22:47
@lisitsynwe also store copy of vectors as we are going to rotate them a little22:48
@lisitsynnext, we project random vectors to each component22:48
@lisitsynthen we modify vectors so that we get them orthogonal to random vector22:49
@lisitsynI don't know what is the meaning but that's the thing I notice to be similar to G-S22:49
@wikingoh shit i haven't even realized that we have Marsaglia polar method for std normal distrib22:49
@lisitsynand finally we rotate G22:50
@lisitsynactually what we see here is just unrolled Givens (rotation) matrix multiplication22:50
@lisitsynwhat we have found that way is a transformation matrix that aligns it22:50
@lisitsynby it I mean P22:51
@lisitsynwiking: yes of course22:51
@lisitsynI even remember who did that22:51
hushellHey guys, what's blame list?22:51
@wikinglisitsyn: u?22:51
@lisitsynno22:51
@wikinglisitsyn: HeikoS  i guess22:51
@lisitsynwiking: no!22:51
@lisitsynjustin patera22:52
@wikingi should do ziggurat22:52
@lisitsynhushell: just a list of emails who could be blamed for breakin things22:52
@wikingwhere's justin?22:52
@lisitsynwiking: he didn't get to gsoc 201122:53
@lisitsynso he is not with shogun :)22:53
@wiking:(22:53
hushellI saw the merge from my branch failed in travis :( and been blamed22:54
@lisitsynhushell: oh yeah we have everything failing22:54
@lisitsyn:D22:54
hushell:D what's wrong here? anyway to fix?22:54
@lisitsynwe need stalin or so I don't know22:55
@wikingWE NEED LENIN!22:55
@wikingwhere's good old comrad lenin22:55
@lisitsynkim jong il22:55
@wikingmmm we need a lenin bot!22:55
-!- iglesiasg [d58f32bd@gateway/web/freenode/ip.213.143.50.189] has joined #shogun22:55
@wikingmmm what's a good bot nowadays?22:56
@lisitsynwiking: lenin? in red house right away from kremlin22:56
@lisitsynlying dead22:56
@lisitsyn:D22:56
@wikingeggdrop still any good?22:56
-!- mode/#shogun [+o iglesiasg] by ChanServ22:56
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun22:56
@lisitsynwiking: no idea!22:56
@lisitsyneggdrop is something from my deep youth22:56
@wikinghahahah22:56
@wikingand u are like younger than me about 10 years22:57
@wikingso imagine that eggdrop is from my youth as well22:57
@wiking;P22:57
@lisitsynI think 822:57
@lisitsyn:D22:57
@wikingInitial release December 199322:57
@wiking:D22:57
@wikingniiiize22:57
@wikingof course i need tcl for it :DDD22:57
@wikingno worries it's already building22:57
@wikinglenin bot is coming!22:58
@sonney2klisitsyn, recall konrads bot here22:58
@lisitsynwiking: right two years after my born22:58
@sonney2kthat was nice22:58
@sonney2kback in the days...22:58
@lisitsynsonney2k: yeah cool bot22:58
@wikingshould we call him lenin or vlad?22:58
@sonney2kwhat was her name?22:58
@sonney2kbetty?22:58
@lisitsynsonney2k: bettyboo22:58
@wikingsonney2k: sooooooo random on (0,1)?22:58
@wikingyes yes?22:58
@sonney2kwiking, you mean closed interval right?22:59
@wikingsonney2k: i mean open22:59
@lisitsynshit lenin says22:59
@sonney2kwiking, btw you can provide another non-closed interval function22:59
@wikingi can do [0,1) and (0,1]22:59
@sonney2kand then we can gradually switch22:59
@wikingbut not fast [0,1]22:59
@sonney2kdon't make this change too complex22:59
@lisitsynwiking: just return 0 sometimes23:00
@lisitsynthat's ok!23:00
@wiking'sometimes'23:00
@wiking:D23:00
@lisitsynyes!23:00
@lisitsynfrom time to time come on23:00
@lisitsynthat's pure random23:00
@sonney2kahh nice my vpn to the states is good23:00
vgorbatilisitsyn: I am just curious if I am right in this logical sequence: we can rotate only around a basis of vectors; if, say, we want to align with 1 principal vector in 3D space - we need 2 more vectors that will be perpendicular to that one -> but that will be 2 other principal components -> there's no way you can align with some subset of principal components without getting the other?23:00
@wikingif (static counter > 123)  counter =0 return -23:00
@wikingreturn 023:00
@wikingright? :)23:01
@sonney2kwiking, I don't know you :P23:01
@wikingsonney2k: and that's actually true :D23:01
@lisitsynwiking: yes that's pure random thing23:01
@lisitsynvgorbati: as I understand it is not *exact* align23:01
@lisitsynvgorbati: one remark here23:02
@lisitsynerich maria remarque23:02
@lisitsyn:D23:02
@lisitsynokay23:02
vgorbatilisitsyn: :D23:02
@lisitsynI mean23:02
@lisitsynthe basis is not made of target_dim vectors23:03
@lisitsynit is made of vectors itself23:03
@sonney2kwiking, did you ever try seafile?23:03
@wikingsonney2k: what'sthat23:04
@lisitsynvgorbati: these vectors are just project so that we get them non-orthogonal23:04
@lisitsynprojected*23:04
@lisitsynsorry, cont'd non-orthogonal to principal components23:05
@wikingsonney2k: i dont want to actually create more api for cmath23:05
@wikingsonney2k: i mean now random calls are still there but actually they shouldn't be... those should be marked deprecated23:05
@wikingand everybody shoudl start calling the global sg_rand directly23:06
@sonney2kwiking, just refactor that then23:07
@sonney2kI am very much fine with that23:07
vgorbatilisitsyn: trying to make it clear for me: the algorithm tries to project every data vector so that it won't be orthogonal with any of target_dim principal components?23:07
@lisitsynvgorbati: yes somewhat23:08
@lisitsynvgorbati: it constructs a basis23:08
@lisitsynthat is not orthogonal to principal components (as possible)23:08
@wikingsonney2k: ok first i add ziggzag to Random as well23:08
@lisitsynso when you project your vectors on that basis23:08
@sonney2kwiking, seafile is an opensource dropbox23:08
@wikingi.e. to have a fast gaussian generator23:09
@lisitsynyou get them 'rotated' so that they are not orthogonal to principal components too23:09
@lisitsynI think the idea is that23:09
@lisitsynis that I described I mean :)23:09
@lisitsynsonney2k: what is wrong with das dropboox?23:10
vgorbatilisitsyn: the main question: why?:) Why not just use pure PCA, but for dimension > target_dim23:11
@lisitsynvgorbati: if I knew23:11
@lisitsyn:D23:11
@lisitsynwiking: https://github.com/lisitsyn/tapkee_website/blob/master/src/tapkee/core.clj more crazyness for you I know you like it23:18
@lisitsyn:D23:18
@lisitsynthat's what is being run here http://tapkee.lisitsyn.me/23:19
vgorbatilisitsyn: ok, leaving now, see you later23:19
@wikingnjiha23:19
@wikinglol23:19
@lisitsynvgorbati: alright, see you23:20
-!- vgorbati [d4029f22@gateway/web/freenode/ip.212.2.159.34] has quit [Quit: Page closed]23:20
-!- hushell [43bd6474@gateway/web/freenode/ip.67.189.100.116] has quit []23:20
@lisitsynwiking: clojure in action :D23:20
@wikingyeps23:20
@lisitsynschlafen!23:31
@sonney2klisitsyn, man if you would spend half the time on shogun ...23:33
@sonney2kbeautiful examples - can we steal them?23:33
@lisitsynsonney2k: I do not mind doing them for shougn *once I am done with f&S*$#$ paper*23:34
@sonney2kok so in 10 years23:34
@lisitsynsonney2k: chris says he has just a few issues left23:35
@lisitsyn:D23:35
@lisitsynsonney2k: I think I could even submit tomorrow23:36
@lisitsynsonney2k: what slows down is that we have iterations over paper (like one per week due to chris being busy)23:36
@lisitsynsonney2k: as for examples - we could put them live easily I think23:37
@lisitsynjust copy images, .js-es and jsons23:38
@sonney2klisitsyn, but they are static right?23:39
@sonney2knot live?23:39
@lisitsynsonney2k: lets just ask foulwall to merge it from my sources23:39
@lisitsynsonney2k: what do you mean by static/live?23:39
@sonney2knot computed on-the-fly23:40
@lisitsynahh no23:40
@sonney2kthe embedding I mean23:40
@sonney2kthat is sth we need23:40
@lisitsynsonney2k: easy - I just host on heroku which is troublesome for that23:40
@lisitsynsonney2k: just give them foulwall I think :)23:41
@lisitsynsonney2k: mnist embedding is ~1m btw23:45
@lisitsynnot reasonable for web example23:45
@lisitsynI don't expect it to be any faster with t-sne23:45
@sonney2klisitsyn, we don't need all 60k examples23:48
@lisitsynsonney2k: ohh 60k is much more23:51
@lisitsynI used some subset, 2000 may be23:51
@lisitsynsonney2k: once it callback'ed more I could add a sparse callback which would be nice for mnist23:52
@sonney2khmmh pretty slow23:53
@sonney2kthen we have to subsample even more23:53
@sonney2kanyway bed time23:53
van51sonney2k: can you wait for a couple of questions? :p23:56
van51anyway, I'll shoot and you can answer in the morning. no rush23:57
van51about the conversion in comb features/kernel, there are some methods like "get_first_kernel (CListElement *&current)" that are dependant on the CList representation. should they be removed?23:58
van51or should they just do what they are supposed to do and be marked as deprecated or something?23:59
--- Log closed Thu May 09 00:00:37 2013
« Tuesday, 2013-05-07 Index Thursday, 2013-05-09 »

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