Open in new window / Try shogun cloud
--- Log opened Sat Jul 20 00:00:35 2013
--- Day changed Sat Jul 20 2013
thoralfHey iglesiasg00:00
@lisitsyniglesiasg: okay I succeed with dvi!00:04
@iglesiasglisitsyn: good job!00:04
@lisitsyniglesiasg: I re-rendered pics00:05
@lisitsynso they are now pure eps with no ugly convertation00:05
@iglesiasglisitsyn: aham! so the problem was passing them from pdf to ps?00:05
@lisitsynno idea but pdfs are not encouraged as we can see00:05
@iglesiasgI understand00:05
@lisitsyniglesiasg: dvi looks very ugly but that's ok I guess00:08
@lisitsyniglesiasg: yeah ps looks much better00:09
@lisitsynthough we have some issue with citations00:10
@iglesiasglisitsyn: my viewers get stuck trying to open it00:13
@lisitsyniglesiasg: hmm it is ok here00:18
@iglesiasglisitsyn: damn OS00:20
@lisitsyniglesiasg: alright I resolved it!00:21
@lisitsyniglesiasg: okay tomorrow's morning I'll scan things and then we can send00:21
@iglesiasglisitsyn: that sounds pretty good00:22
-!- HeikoS [] has quit [Quit: Leaving.]00:47
thoralfsonney2k: Ping?01:23
@iglesiasggood night people01:47
-!- iglesiasg [] has quit [Quit: Leaving]01:54
-!- nube [~rho@] has joined #shogun02:45
-!- nube [~rho@] has quit [Client Quit]02:49
shogun-buildbotbuild #398 of nightly_all is complete: Failure [failed compile]  Build details are at
-!- foulwall [~user@2001:da8:215:503:9007:bf2c:83a4:58f3] has joined #shogun03:21
-!- Yanglittle [b74040fc@gateway/web/freenode/ip.] has joined #shogun03:39
Yanglittleexcuse me.03:41
-!- thoralf [] has quit [Quit: Konversation terminated!]03:48
-!- Yanglittle [b74040fc@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]04:11
shogun-buildbotbuild #463 of nightly_default is complete: Failure [failed test]  Build details are at
-!- foulwall` [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has joined #shogun04:21
-!- foulwall [~user@2001:da8:215:503:9007:bf2c:83a4:58f3] has quit [Ping timeout: 245 seconds]04:23
-!- pickle27 [] has quit [Quit: Leaving]06:45
-!- foulwall` [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has quit [Read error: Connection reset by peer]07:28
-!- foulwall` [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has joined #shogun07:50
-!- gsomix_ [~gsomix@] has joined #shogun08:43
-!- gsomix [~gsomix@] has quit [Ping timeout: 264 seconds]08:47
-!- foulwall` [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has quit [Remote host closed the connection]09:01
-!- foulwall [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has joined #shogun09:01
-!- thoralf_ [] has joined #shogun10:33
thoralf_Good morning.10:34
-!- thoralf_ is now known as thoralf10:34
-!- lambday [67157d37@gateway/web/freenode/ip.] has joined #shogun10:45
-!- thoralf_ [] has joined #shogun10:48
-!- zxtx_ [] has joined #shogun10:48
-!- gsomix__ [~gsomix@] has joined #shogun10:48
-!- Netsplit *.net <-> *.split quits: thoralf, gsomix_, zxtx10:54
-!- Netsplit *.net <-> *.split quits: @sonney2k, @lisitsyn, gsomix__11:06
-!- Netsplit over, joins: gsomix__, @lisitsyn, @sonney2k11:07
-!- sonney2k_ [] has joined #shogun11:18
-!- Netsplit *.net <-> *.split quits: @sonney2k11:24
-!- iglesiasg [] has joined #shogun12:02
-!- mode/#shogun [+o iglesiasg] by ChanServ12:02
@iglesiasghi guys12:02
-!- thoralf_ [] has quit [Quit: Konversation terminated!]12:34
-!- sonney2k_ is now known as sonney2k12:38
-!- mode/#shogun [+o sonney2k] by ChanServ12:38
@sonney2kfoulwall, gsomix any Q's ?12:38
-!- van51 [] has joined #shogun12:53
-!- gsomix__ [~gsomix@] has quit [Remote host closed the connection]12:56
-!- Netsplit *.net <-> *.split quits: van5113:02
-!- Netsplit over, joins: van5113:03
-!- foulwall [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has quit [Ping timeout: 245 seconds]13:10
-!- shogun-notifier- [] has joined #shogun13:15
shogun-notifier-shogun: Soeren Sonnenburg :develop * bf51d1d / doc/ipython-notebooks/classification/svm_binary_classification.ipynb:
shogun-notifier-shogun: add beautified svm notebook from talk13:15
shogun-buildbotbuild #1484 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>13:18
-!- nube [~rho@] has joined #shogun13:23
-!- nube [~rho@] has quit [Client Quit]13:25
-!- travis-ci [] has joined #shogun13:43
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []13:43
van51sonney2k: hey13:44
-!- foulwall [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has joined #shogun13:47
-!- nube [~rho@] has joined #shogun13:47
foulwallsonney2k: cant came up with a good solution for Ai.py13:47
-!- Avasz [~Avash@unaffiliated/avasz] has joined #shogun14:10
-!- gsomix [~gsomix@] has joined #shogun14:18
-!- foulwall [~user@2001:da8:215:503:a8e5:5b82:ed61:9640] has quit [Remote host closed the connection]14:54
-!- van51 [] has quit [Quit: Leaving.]15:20
-!- nube [~rho@] has quit [Quit: Leaving.]16:09
-!- shogun-notifier- [] has quit [Quit: transmission timeout]16:15
-!- lambday [67157d37@gateway/web/freenode/ip.] has quit []16:22
-!- thoralf [] has joined #shogun17:07
thoralfsonney2k: I just read your comment about the check-examples PR.  If you think the changes are "okay", then I could apply them to the other interfaces as well.17:10
-!- Avasz [~Avash@unaffiliated/avasz] has left #shogun ["ban-kai"]17:25
-!- thoralf [] has quit [Quit: Konversation terminated!]17:27
-!- foulwall [~user@2001:da8:215:c252:4d94:a3ff:4e13:e516] has joined #shogun17:45
-!- iglesiasg [] has quit [Quit: Leaving]18:15
-!- gsomix [~gsomix@] has quit [Ping timeout: 248 seconds]18:52
-!- foulwall [~user@2001:da8:215:c252:4d94:a3ff:4e13:e516] has quit [Remote host closed the connection]19:11
-!- gsomix [~gsomix@] has joined #shogun19:44
-!- gsomix [~gsomix@] has quit [Read error: Connection reset by peer]19:48
-!- pickle27 [] has joined #shogun20:14
@sonney2klisitsyn, do you know how much memory using an object inside another object takes?20:37
@lisitsynsonney2k: what do you mean by using?20:37
@lisitsynlike member?20:37
@sonney2klisitsyn, yes20:38
@lisitsynsonney2k: with a pointer or?20:38
@lisitsynsonney2k: iirc just sizeof(...) but I am not sure what you mean then20:42
@sonney2klisitsyn, I will commit and then show you20:43
-!- gsomix [~gsomix@] has joined #shogun21:05
gsomixgood evening21:05
gsomixthis day was hard - many work at home21:05
gsomixso, unit-tests time now. :)21:06
-!- hoijui [] has joined #shogun21:34
-!- shogun-notifier- [] has joined #shogun21:36
shogun-notifier-shogun: Soeren Sonnenburg :develop * 209a17c / src/shogun/ (12 files):
shogun-notifier-shogun: introduce CLock object21:36
shogun-notifier-shogun: - do pthread etc based locking using CLock from CMap, SGObject, Random21:36
@sonney2klisitsyn, finally21:38
@sonney2kplease have a look at SGReferencedData's21:38
@sonney2kRefCount type21:38
@sonney2kI have a CLock class in there now21:38
@sonney2kif we could do this w/o the lock object - would be even better21:38
@sonney2kif a type like atomic int in java did exist we should rather use that21:39
@lisitsynsonney2k: why to avoid it?21:39
@lisitsynwell C++11 has atomic21:39
@lisitsynsonney2k: just check sizeof(CLock)21:40
@lisitsynI wouldn't expect more than 4 or 8 IDK21:41
@sonney2klisitsyn, what is the size of atomic<int> ? issue is that this is inside of SGReferencedData21:41
@sonney2kand every byte counts there21:42
@lisitsynI wouldn't count bytes here21:42
@lisitsynit is just a few bytes that's ok I think21:42
@lisitsynI don't know about atomic but it should be quite similar to what we have now21:43
@sonney2klisitsyn, if we have *many* 2d-vectors of bytes21:43
@sonney2kwe have sth like factor 8 overhead21:44
@sonney2kso any byte counts21:44
@lisitsynsonney2k: ain't 2d a bit strange case/21:44
@lisitsynsonney2k: anyway why did you hack it btw?21:48
@lisitsynI mean any trouble with it?21:48
shogun-buildbotbuild #1485 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>21:57
-!- travis-ci [] has joined #shogun22:00
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []22:00
@sonney2klisitsyn, I want to get rid of pthread in headers22:10
@sonney2ksame holds for other libs btw22:11
@lisitsynsonney2k: good idea22:11
@sonney2kbesides it is much cleaner code to have a single Lock class than all the time doing pthread_init / destroy lock /unlick22:11
@sonney2klisitsyn, an atomic int seems like the solution to this - it requires just 4 bytes (like an int)22:23
@sonney2kno need for any memory overhead22:23
@sonney2kat least on x86_6422:23
@lisitsynsonney2k: uh that's interesting22:23
@lisitsynit is a structure still right?22:24
@lisitsynI guess they have no virtual things here (even destructor)22:24
@lisitsynthis makes it easier to optimize everything out22:24
@sonney2klisitsyn, try
@sonney2klisitsyn, yeah nothing can be virtual in there22:24
@sonney2kproblem is I have to compile with -std=c++1122:25
@lisitsynhah well22:25
@lisitsynyou can steal atomic22:25
@lisitsynsonney2k: the output is22:26
@lisitsyn0 422:26
@lisitsynjust out of curiosity22:26
@sonney2kso 0 overhead22:27
@sonney2kthat is actually pretty neat22:27
@lisitsynsonney2k: structure with only one int field22:28
@lisitsynis 4 bytes too22:28
@sonney2kwell ok there willb e sync overhead22:28
@sonney2kbut not memory22:28
@sonney2kwhich is what counts here22:28
@lisitsynsonney2k: this load method should go as asm op directly22:29
@lisitsynI mean that's atomic operation in x8622:29
@lisitsynso it is barely any overhead from any point of view22:29
@sonney2kthe processor has to do some syncing underneath22:30
@lisitsynI mean it can't be any better22:31
@sonney2kyes sure22:33
@sonney2kit is like 5 times slower than a normal int22:33
@sonney2ktakes 0.2 vs 1 sec here22:33
@lisitsynsonney2k: 5 times is rather surprisingly good22:34
@sonney2klisitsyn, hmmhh I would love to use it...22:35
@lisitsynsonney2k: steal it22:36
@lisitsynsonney2k: I just checked the sources22:36
@lisitsynsonney2k: it is a bit C++11y22:36
@lisitsynbut we may patch it22:37
@lisitsynI'd actually use C++11 already haha22:37
@sonney2klisitsyn, do we even have it on our buildbots?22:37
@lisitsynsonney2k: what gcc version?22:37
@lisitsyn4.6 has it22:37
@sonney2kno idea22:37
@sonney2khmmhh 4.4 is the oldes22:38
@lisitsyn4.4 has no atomic I think22:38
@sonney2kredhat also has only 4.422:39
@sonney2klisitsyn, I think we have to do a fallback thing22:42
@sonney2klisitsyn, sth like if c++11 is available22:42
@sonney2kwe use atomic22:43
@sonney2kotherwise CLock + whatever overhead22:43
@lisitsynsonney2k: yeah as usual22:43
@lisitsynsonney2k: sig?22:44
-!- hoijui [] has quit [Quit: Leaving]22:45
@lisitsynwhat is the type you are talking about?22:45
@sonney2kTo avoid uncertainty about interrupting access to a variable, you can use a particular data type for which access is always atomic: sig_atomic_t. Reading and writing this data type is guaranteed to happen in a single instruction, so there's no way for a handler to run "in the middle" of an access.22:46
@sonney2kThe type sig_atomic_t is always an integer data type, but which one it is, and how many bits it contains, may vary from machine to machine.22:46
@lisitsynoh didn't know about that22:47
@lisitsynsonney2k: that's compiler level stuff already as I can see22:47
@lisitsynso strictly forces to use some -std=..22:48
@sonney2kI would prefer the c++11 one though22:50
@lisitsynsonney2k: yeah we are in C++ not C22:52
@sonney2klisitsyn, so I would add at test for C++11 in configure and then use it in my Lock class and also SGReferencedData22:54
@lisitsynsonney2k: oh that's only beginning then :D22:55
@lisitsynwe have some more places to use 1122:55
@sonney2katomic you mean?22:55
@sonney2kor C++11?22:55
@lisitsynno other features22:56
@lisitsynwell I'd have to recall what ideas I had22:56
@sonney2kand recall that we need to be working w/o C++11 still :/22:56
@lisitsynsonney2k: nullptr is a nice thing22:56
@lisitsynsonney2k: shared_ptr again22:57
@sonney2kwhat gain gives us nullptr?22:58
@lisitsynsonney2k: just safety22:58
@sonney2kwrt what?22:58
@lisitsynsonney2k: you can't assign a pointer with 022:59
@sonney2kI mean the compiler complains already if there are 0 vs NULL mixups22:59
@lisitsynwell NULL is still 0 right?22:59
@lisitsynnullptr is just a structure22:59
@lisitsynso harder to put something wrong22:59
@sonney2klisitsyn, but will then sth like if (!ptr) { } be valid still23:00
@lisitsynsonney2k: yes sure23:00
@lisitsynit is impossible to change23:00
@lisitsynsonney2k: just assignment issue23:00
@sonney2kbtw I read the remaining nips papers today23:01
@sonney2kall crap23:01
@lisitsynsonney2k: ahh one more thing23:01
@sonney2kbut now I am to tired to write the reviews23:01
@lisitsynsonney2k: nullptr is better for functions that accept pointers23:01
@sonney2kand for some reason my kids don't want to sleep early23:01
@lisitsynlike f(int*) and f(int)23:01
@sonney2kahh ok23:02
@lisitsynsonney2k: have you accepted a paper?23:02
@sonney2kdon't ask me how many I accepted23:04
@sonney2knot many...23:04
@sonney2kI did reviews for 40 journals/conferences since I am doing this23:04
@sonney2kso I am totally lost23:04
@lisitsynsonney2k: well I think I understand why it is so23:05
@lisitsynthe progress is usually slight23:05
@sonney2kthat is not it but papers have some flaws23:05
@sonney2kas in experiments hide sth that is troublesome23:06
@sonney2kor are wrong23:06
@lisitsynsonney2k: what is this year nips about?23:06
@lisitsynstill deep learning? :D23:06
@sonney2kheh I guess so23:07
@sonney2kthis trend is there since 2009 or so23:07
@sonney2kand I remember that LeCun's deep learning workshop got rejected back then23:07
@lisitsynsonney2k: I am curious what is next23:07
@sonney2kkernel machines :D23:08
@sonney2kwell NN's are not exactly new aren't they?23:11
@lisitsynsonney2k: haha yes a bit old23:11
@sonney2ksleep time23:11
@lisitsynpickle27: hey just noticed you are here23:11
pickle27yeah I actually just got up from a nap23:12
pickle27left the comp on23:12
pickle27lisitsyn: did you see my PR what do you think of it now?23:13
@lisitsynpickle27: ohh sorry let me check23:13
@lisitsynpickle27: got it! so diago is diag orthogonalization?23:14
pickle27I didn'23:14
pickle27t really get that until Andres pointed it out23:14
-!- lisitsyn1 [] has joined #shogun23:17
lisitsyn1hmm something got wrong23:17
-!- lisitsyn [] has quit [Read error: No route to host]23:17
-!- lisitsyn1 is now known as lisitsyn23:17
lisitsynpickle27: did you receive my last message?23:18
lisitsynon LeftRotStack23:18
lisitsynpickle27: could you please rename that stuff too23:18
pickle27with underscore?23:18
lisitsynjust to follow the naming23:18
pickle27sounds good23:18
pickle27I wasn't sure if you intended that or not23:18
pickle27lisitsyn: any other comments?23:28
lisitsynpickle27: no lets merge if it is done23:28
pickle27its done!23:28
lisitsyngood thanks23:28
shogun-notifier-shogun: Kevin :develop * 2fcc7b6 / src/shogun/ (12 files):
shogun-notifier-shogun: added the proper JointDiago method for Jade, cleaned up the includes in AJD and fixed a typo23:28
shogun-notifier-shogun: Kevin :develop * 22dd91a / / (2 files):
shogun-notifier-shogun: fixed Jade unit test and made some updates to JointDiago23:28
shogun-notifier-shogun: Kevin :develop * 26be32a / src/shogun/mathematics/ajd/UWedge.cpp,src/shogun/mathematics/ajd/UWedge.h:
shogun-notifier-shogun: changed eps for UWedge and sorted the eigenvectors for V023:29
shogun-notifier-shogun: Kevin :develop * b37f799 / src/shogun/ (5 files):
shogun-notifier-shogun: updated name of JointDiago to JADiagOrth23:29
shogun-notifier-shogun: Kevin :develop * fa9f83b / src/shogun/mathematics/ajd/JADiagOrth.cpp:
shogun-notifier-shogun: renamed JADiagOrth inner functions23:29
shogun-notifier-shogun: Sergey Lisitsyn :develop * 1d76616 / / (13 files):
shogun-notifier-shogun: Merge pull request #1269 from pickle27/develop23:29
shogun-notifier-shogun: added the proper JointDiago method for Jade, cleaned up the includes in ...23:29
lisitsynpickle27: about a demo23:29
lisitsynso what are we going to implement?23:30
pickle27lisitsyn: yeah Im planning to start on the audio demo this week23:30
pickle27should be fairly similar to the example in python_modular/graphical only with audio and playback23:30
lisitsynpickle27: I'll try to check how to do that with d3.js23:30
pickle27then I'd like to go web but I need to look at foulwalls stuff to see what to do23:30
lisitsynpickle27: would be nice to see some graph23:30
lisitsynhow does it called23:31
pickle27yeah sort of like what sound cloud has23:31
pickle27I have no idea23:31
lisitsynyes yes23:31
lisitsynthere should be some name23:31
pickle27I was going to google js+soundcloud+playback23:31
pickle27hmm maybe I could even pull songs from sound cloud23:33
lisitsynwhy not23:33
shogun-buildbotbuild #1486 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Kevin <>23:34
pickle27anyways this week I'll work on an offline audio example and then start looking into making it web the later in the week and next23:40
pickle27also hmmm23:40
lisitsynpickle27: just plot it some way you find convenient23:40
lisitsynlater we can convert23:40
pickle27I can not for the life of me figure why the Jade unit test fails23:41
pickle27I run the exact same code not in the unit test framework and it runs23:41
pickle27I was hoping it would clear up on the build bot23:41
lisitsynpickle27: well everything is broken with clone23:41
lisitsynI have to check again23:41
pickle27I will keep tabs on it23:42
pickle27but yeah I wouldn't mind waiting to fix it until everything else settles unit test wise23:42
lisitsynthat'd be better23:42
pickle27anyways Im out, catch you later!23:43
lisitsynsee you23:44
-!- pickle27 [] has quit [Quit: Leaving]23:44
shogun-buildbotbuild #1487 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Kevin <>23:47
shogun-buildbotbuild #1488 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>, Kevin <>23:48
lisitsynhaha build 14 8823:52
--- Log closed Sun Jul 21 00:00:40 2013