Open in new window / Try shogun cloud
--- Log opened Thu Aug 29 00:00:35 2013
-!- hushell [] has quit [Ping timeout: 245 seconds]00:01
-!- travis-ci [] has joined #shogun00:26
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:
-!- travis-ci [] has left #shogun []00:26
-!- hushell [] has joined #shogun00:41
-!- iglesiasg [] has quit [Quit: Ex-Chat]00:45
-!- travis-ci [] has joined #shogun00:54
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:
-!- travis-ci [] has left #shogun []00:54
-!- pickle27 [] has quit [Quit: Leaving.]01:41
-!- foulwall [~user@2001:da8:215:6901:a071:8f2f:8c28:627a] has joined #shogun06:44
foulwallping sonne|work sonne|osx sonney2k06:52
-!- sonne|osx [] has quit [Quit: sonne|osx]07:16
-!- hoijui [] has joined #shogun07:19
-!- sonne|osx [~sonne@] has joined #shogun08:08
sonne|osxfoulwall: hey08:08
sonne|osxfoulwall: please don't always hide from us - be in IRC :)08:08
foulwallsonne|osx, my irc client did not do auto reconnect:(08:10
sonne|osxfoulwall: well use one that does :)08:10
sonne|osxfoulwall: anyway08:11
sonne|osxfoulwall: do you remember the multi class demo from the beginning?08:11
sonne|osxfoulwall: we just realized that this is still using the old code david did. Please convert it to your framework and get rid of davids old code08:11
sonne|osxgsomix__: moin moin!08:13
foulwallsonne|osx, ok, I'll rewrite them this afternoon, but I'll finish the mldata downloading first. I'll send a pull request on mldata downloader this afternoon.08:13
gsomix__sonne|osx, hey.08:13
sonne|osxfoulwall: ok thanks!08:14
sonne|osxgsomix__: now that the ascii stuff works what would you want to do next?08:14
gsomix__sonne|osx, protobuf08:15
sonne|osxgsomix__: ok so how much do you know already about it?08:16
sonne|osxgsomix__: maybe I just explain quickly08:16
sonne|osxgsomix__: you need to define a .proto file08:16
sonne|osxthat is similar to a struct in C08:16
gsomix__sonne|osx, will be fine. please, explain.08:16
sonne|osxwhere you define what is to be saved08:16
sonne|osxthen this file can be compiled into C code08:16
sonne|osxand for saving you will then fill the structs provided in this C code08:17
sonne|osxand then there is some function that gives you a byte array08:17
sonne|osxthat you can then dump to disc08:17
sonne|osxas you can see the problem is that you need basically everything extra in memory08:18
sonne|osxso it is not meant to store e.g. a full SGMatrix at once08:18
sonne|osxbut rather chunks of the matrix08:18
sonne|osxlets start with a simple protobuf file for a vector08:19
sonne|osxyou have 2 fields (for now)08:20
sonne|osx{ int64 length = 1; repeated double vector = 2; }08:21
sonne|osxthe 1 and 2 here are identifiers08:21
sonne|osxyou then get the protobuf tools08:21
sonne|osxand compile that file with protoc08:21
sonne|osxand then just try for once to fill in the data08:22
sonne|osxgsomix__: after that is done you are basically ready and we will define protobuf files for all of what CFile supports (yes for all datatypes like int etc).08:22
gsomix__sonney2k, ok, got it.08:23
sonne|osxgsomix__: later we can make things more efficient08:23
sonne|osxby saving multiple chunks08:23
sonne|osxi.e. instead of saving the whole vector08:23
sonne|osxyou save say only a 1MB chunk of the vector08:23
sonne|osxand then append another etc08:23
sonne|osxfor that to work you need a header08:24
sonne|osxthat says how big the chunk is08:24
sonne|osxand another header saying how big the vector really is08:24
sonne|osxand well it makes sense to have another header saying which format version this is08:24
sonne|osxso we can change things later if we want08:24
sonne|osxgsomix__: ok?08:25
gsomix__sonne|osx, ok. need little time for comprehension :)08:26
gsomix__simple protobuf file for vector vector first08:26
sonne|osxwell ask whe you have Qs08:26
sonne|osxoki gtg be back soon08:26
-!- sonne|osx [~sonne@] has quit [Quit: sonne|osx]08:26
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has joined #shogun08:32
-!- foulwall [~user@2001:da8:215:6901:a071:8f2f:8c28:627a] has quit [Ping timeout: 264 seconds]08:35
sonne|workgsomix__: RE08:40
gsomix__sonne|work, ?08:40
sonne|workgsomix__: RE == back08:41
sonne|workgsomix__: so questions?08:41
gsomix__many, many sonne's...08:41
gsomix__sonne|work, not yet08:41
-!- hoijui [] has quit [Quit: Leaving]08:44
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has quit [Ping timeout: 264 seconds]09:00
-!- shogun-notifier- [] has joined #shogun09:01
shogun-notifier-shogun: hushell :develop * 578f0f9 / / (19 files):
shogun-notifier-shogun: fix initialization bugs of factorgraph stuffs09:01
shogun-notifier-shogun: Fernando Iglesias :develop * fe9d502 / / (19 files):
shogun-notifier-shogun: Merge pull request #1476 from hushell/dev_bugs09:01
shogun-notifier-shogun: fix initialization bugs of factorgraph stuffs09:01
shogun-buildbotbuild #1559 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: hushell <hushell@hushell-U510.(none)>09:15
shogun-buildbotbuild #1560 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Fernando Iglesias <>09:20
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has joined #shogun09:35
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has left #shogun []09:36
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has joined #shogun09:36
shogun-buildbotbuild #1644 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at
-!- lisitsyn [] has quit [Quit: Leaving.]09:46
shogun-buildbotbuild #1023 of rpm1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: hushell <hushell@hushell-U510.(none)>09:49
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:8181:8528:e5a1:bac7] has joined #shogun09:51
-!- mode/#shogun [+o iglesiasg] by ChanServ09:51
@iglesiasggood morning guys09:51
shogun-buildbotbuild #1344 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: hushell <hushell@hushell-U510.(none)>09:53
shogun-buildbotbuild #1645 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at  blamelist: Fernando Iglesias <>10:02
-!- van51 [] has joined #shogun10:06
-!- foulwall` [~user@2001:da8:215:6901:47b:de83:2dac:7945] has quit [Ping timeout: 264 seconds]10:10
shogun-buildbotbuild #1024 of rpm1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Fernando Iglesias <>10:31
shogun-buildbotbuild #1345 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Fernando Iglesias <>10:39
-!- lisitsyn [] has joined #shogun10:41
-!- HeikoS [] has joined #shogun11:25
-!- mode/#shogun [+o HeikoS] by ChanServ11:25
@HeikoSsonne|work: around?11:32
sonne|workHeikoS: y11:32
@HeikoSsonne|work: so if we do a connection flight11:32
@HeikoSits as cheap as 534 return p.P.11:33
@HeikoSover minneapolis11:33
@HeikoSbut takes quite a bit longer11:33
@HeikoSjust found another cheap one11:34
lisitsynHeikoS: what's the route for you?11:34
@HeikoSlisitsyn: london -> SFO11:35
lisitsynah directly11:35
@HeikoSsonne|work: so there is one for 544 pounds directly london sfo11:35
lisitsynI will have KUF -> SVO -> JFK -> SFO11:35
@HeikoSLHR->SFO 1035->1335        SFO->LHR 13:00->7:10       for 544 each11:35
@HeikoSlisitsyn: oh man11:36
lisitsynHeikoS: but that's just 1000$11:36
lisitsynboth back and forth11:36
@HeikoSlisitsyn: stressfull :)11:36
@HeikoSah cool11:36
@HeikoSthats good11:36
lisitsynyeah really cheap11:36
lisitsynI guess moscow -> sfo is way too long way11:36
sonne|workHeikoS: don't get it ... what is the total? FLight I looked up where about 700-800 back and forth :)11:36
@HeikoSsonne|work: so if you would be in london LHR by 9 everythign is fine11:37
@HeikoSsonne|work: 544 return each11:37
lisitsynHeikoS: do you want to share a flight?11:37
sonne|workHeikoS: return means LHR->SFO & SFO->LHR?11:37
@HeikoSsonney2k: yes, that are 636 euros11:37
@HeikoSsonne|work: a flight to london should be cheap, less than 10011:38
sonne|workok so I would maybe pay 300 or so to get to LHR and back11:38
@HeikoSsonne|work: no thats much less11:39
@HeikoSI fly to Germany often, its usually under 100 euros for two people11:39
lisitsynHeikoS: lucky you11:39
@HeikoSsonne|work: so go together or seperate?11:40
@HeikoSsonne|work: also fine with seperate that might be cheaper11:40
lisitsynI go to us embassy *by train* and it is more than 100euros for me alone11:40
@HeikoSlisitsyn: oh man sweet11:40
lisitsynHeikoS: not really :) that's too much11:41
lisitsynHeikoS: comparing to your london - germany trips for 100 euros11:41
@HeikoSlisitsyn: for two :D11:41
lisitsynHeikoS: unhealthy prices11:41
@HeikoSlisitsyn: but keep in mind its very close11:41
@HeikoSby car would be just 5 hours11:41
sonne|workHeikoS: well book this flight, give me the details and I try to get the same one!11:41
lisitsynHeikoS: hmm true11:42
sonne|workHeikoS: it is friday to friday right?11:42
@HeikoSsonne|work: I will send you the dates11:42
sonne|workHeikoS: I cannot go earlier (no more holidays)11:42
@HeikoSsonne|work: not sure what you mean, shall I book for me only?11:42
sonne|workHeikoS: ???11:42
sonne|workHeikoS: you book for you11:43
sonne|workfrom fr -> fr11:43
sonne|workand then I book the same flight when I have the details11:43
@HeikoSI see,11:43
lisitsynHeikoS: when will you arrive?11:43
sonne|worklisitsyn: I hope there will still be seats so you could join aswell11:44
@HeikoSsonne|work: arriving on sat is also fine right? (flight is at 1300)11:44
sonne|workFriday evening is the first pool party & reception11:44
lisitsynsonne|work: oh moscow -> london is not cheap11:44
@HeikoSsonne|work: return sorry11:44
sonne|workHeikoS: these are the events you get to know most people11:45
@HeikoSsonne|work: was a typo I meant return11:45
sonne|workHeikoS: I think when we leave friday evening we will be back on saturday evening11:45
@HeikoSsonne|work: lisitsyn Ill send you a meail with the flight dates soon to double check, and if you give OK I will book11:45
sonne|workyou know kids... 1 week is already very long11:46
lisitsynHeikoS: I think I'll go separate - I checked moscow -> london and it is already way too expensive11:46
sonne|workHeikoS: please in messenger I dont' read mail at work11:46
lisitsynHeikoS: moscow -> new york is better for me11:46
sonne|worklisitsyn: how much is it to go to SFO for you?11:46
lisitsynsonne|work: 1000$11:46
sonne|worklisitsyn: ok about the same11:47
lisitsynsonne|work: that's samara->moscow->new york->san francisco11:47
lisitsynand back11:47
sonne|worklisitsyn: try to be in SFO around 13:30 :)11:47
@HeikoS18 October 2013 11:47
@HeikoSLHR 10:35 -> SFO 13:3511:47
@HeikoSUNITED 90011:47
sonne|workso we can all share the same flight11:47
lisitsynsonne|work: yes that's what I wanted to ask11:47
@HeikoSI will book now11:47
sonne|workerr car11:47
lisitsynsonne|work: friday 13:3011:47
sonne|worklisitsyn: yes11:48
sonne|workwhat heiko just posted11:48
lisitsynok thanks let me find something11:48
@HeikoSand reutnr11:48
@HeikoS25 October 201311:49
@HeikoSSFO 13:00 -> LHR 07:10 (arrives 26 October)11:49
@HeikoSUNITED 90011:49
@HeikoSsorry UNITED 90111:49
@HeikoSI found this on expedia11:50
sonne|workHeikoS: I wouldn't mind to go back later in the day11:50
@HeikoSok checking11:50
sonne|workif there is a flight of course11:50
sonne|worksounds good11:51
@HeikoSarrives 1215 next day sat11:51
@HeikoS18 October 201311:52
@HeikoSLHR 10:35 -> SFO 13:3511:52
@HeikoSUNITED 90011:52
@HeikoS25 October 201311:52
@HeikoSSFO 18:05 -> LHR 12:15 (arrives 26 October)11:52
@HeikoSLufthansa 9379 Operated by UNITED AIRLINES11:52
@HeikoSboth direct11:52
sonne|worklufthansa even?11:52
@HeikoSoperated by united11:52
@HeikoSsame price11:52
@HeikoSexpedia special price11:52
@HeikoSok booking11:53
@HeikoScanellation insurance? :)11:53
lisitsynI can go lufthansa but it is ~40K roubles11:55
lisitsynlike 1300$11:55
@HeikoSok i booked11:56
sonne|workcheapest flight here is txl->fra->sfo with lufthansa11:57
sonne|work696 EUR11:57
@HeikoShey cool I can choose seats11:58
sonne|workHeikoS: found the flight11:58
sonne|work823 with lufhansa over london11:59
@HeikoSfrom berlin?11:59
@HeikoSsounds good11:59
sonne|workyes agreed11:59
@HeikoSIll send you a screen which seat I chose11:59
@HeikoSin both flights12:00
@HeikoSok sweet12:01
sonne|workHeikoS: I cannot book just now. Need to do some logistics with parents, gf, kids first12:01
@HeikoSsonne|work: ok12:01
-!- shogun-notifier- [] has quit [Quit: transmission timeout]12:01
sonne|workHeikoS: email me the details and I will hopefully book tonight12:01
@HeikoSsent to you and lisitsyn12:02
sonne|workHeikoS: kthx12:02
lisitsynHeikoS: thanks12:02
lisitsynHeikoS: sonne|work: I am having troubles with finding a flight that arrives before you12:04
@HeikoSlisitsyn: if you arrive one day earlier you can stay at my place12:04
lisitsynHeikoS: what place?12:05
sonne|worklisitsyn: when would you arrive?12:05
@HeikoSI live in London remember?12:05
lisitsynHeikoS: ahh you mean london12:05
lisitsynsonne|work: most of the flights arrive to SFO evening12:05
lisitsynlet me check london12:05
lisitsynHeikoS: wait wait12:05
lisitsynhow can I stay in london12:05
lisitsynHeikoS: come on they won't let me in haha12:06
@HeikoSlisitsyn: ah those bastards :)12:06
@HeikoSI once slept a night at heatthrow airport12:06
@HeikoSthat was pretty unconfortable :D12:06
sonne|worklisitsyn: exact time?12:07
lisitsynsonne|work: either 19 or 2112:07
sonne|workthat is too late12:07
lisitsynsonne|work: there is a flight by delta that arrives 12:2812:07
sonne|worktake that one12:07
lisitsynbut it is +300$12:07
sonne|workyou would miss the reception12:07
sonne|workor go a day earlier12:08
lisitsynthat's bad12:08
sonne|worklisitsyn: yeah there is a pool party on friday evening12:08
lisitsynsonne|work: oh this one is supposed to have 19hrs waiting in JFK12:08
sonne|workhaha so stranded in JFK - worse than LHR :D12:10
lisitsynwell I can go out12:11
lisitsynI guess12:11
lisitsynbut yes that's too much12:11
lisitsynHeikoS: I can arrive to heathrow by 08:0012:13
@HeikoSlisitsyn: that sounds good12:13
@HeikoSheathrow is not too bad in fact12:13
lisitsynHeikoS: how much your flight cost?12:13
@HeikoS544 return from london12:13
lisitsyn544 pounds?12:13
lisitsynargh that would be quite a lot12:14
@HeikoSlisitsyn:  you should totally move to europe ;)12:14
lisitsyn300 + 300 + 84412:14
lisitsynHeikoS: that's $ cost12:14
@HeikoSalthough travelling home is tough than12:14
@HeikoSlisitsyn: whats the alternative?12:15
lisitsynHeikoS: arrive evening by aeroflot from moscow through new york12:15
lisitsynarriving at 19:3812:15
@HeikoSlisitsyn: what about going one day earlier?12:16
lisitsynHeikoS: yes possible12:16
sonne|worklisitsyn: so then you rent the car and pick us up at the airport :D12:16
lisitsynsonne|work: no driver license12:17
sonne|worklisitsyn: or you just take a cab12:17
lisitsynsonne|work: are u sure 19:39 is already too late?12:17
lisitsynI mean I arrived to the hotel in berlin like ~23 and it was ok12:18
@HeikoSlisitsyn: the problem is the party aleady is over then12:18
@HeikoSyou need some time from the airport etc12:18
@HeikoSmaybe rest12:18
lisitsynHeikoS: ~1hr I guess12:18
lisitsynhmm no12:18
@HeikoShaha :)12:19
@HeikoSas you like12:19
@HeikoSlisitsyn: would be sad if you missed that12:19
@HeikoSits the reason why you go there12:20
@HeikoSso better go one day ealier12:20
lisitsynHeikoS: yeah makes sense12:20
lisitsynHeikoS: well I can arrive late night of 17th12:20
lisitsynand just walk down to the hotel :D12:20
@HeikoSlisitsyn: yep, have a nice sleep and then we already arrive12:20
lisitsynHeikoS: do you think I can check in night of 18th?12:21
lisitsynCheck-In TimeAnytime after 3:00 pm. Valid ID and credit card required.12:21
lisitsynHeikoS: I can't check in night between 17th and 18th12:22
lisitsynHeikoS: Check-In TimeAnytime after 3:00 pm. Valid ID and credit card required.12:23
lisitsynsonne|work: ha there are rooms in wild palms still12:24
lisitsynHeikoS: oh found one12:29
lisitsynHeikoS: arrives 16:2212:29
lisitsynoh shit12:29
lisitsyn23 hrs in JFK12:29
lisitsynHeikoS: bingo!12:33
lisitsynarrives 12:2012:33
lisitsynargh 14 hrs in frankfurt12:34
@HeikoSlisitsyn: ?12:34
lisitsynHeikoS: all the flights that fit the time have some crazy waiting times so far12:36
@HeikoSah shit12:36
lisitsynHeikoS: searching :)12:36
lisitsynHeikoS: okay catch it12:36
lisitsynoh wonderful wonderful12:36
lisitsynKUF -> FRA -> SFO12:37
lisitsynHeikoS: and just one transfer12:38
@HeikoSthat sounds good12:41
@HeikoShow much?12:41
lisitsynHeikoS: 1300$12:42
@HeikoSthats fine12:42
@HeikoSisnt it?12:42
@HeikoShow much do we get again?12:42
lisitsynHeikoS: no idea :)12:42
lisitsynHeikoS: I am ok to cover it anyway12:43
@HeikoSlisitsyn: should work out fine though12:43
lisitsynHeikoS: I have no idea how does it work12:43
lisitsynand not sure it is easy to transfer money to ?ussia :)12:44
lisitsynso in the worst case I pay by myself but I am ok12:44
lisitsynlunch time12:44
@HeikoSenjoy :D12:48
lisitsynHeikoS: ok then I'll buy it once I get visa13:01
@HeikoSlisitsyn: I have a problem with swig can you help me?13:01
lisitsynHeikoS: for sure13:01
@HeikoSbasically first question: how to do templates?13:01
@HeikoSI mean RealBlabla13:01
@HeikoSI did this13:02
@HeikoS#ifdef USE_FLOAT6413:02
@HeikoS    %template(RealOperatorFunction) COperatorFunction<float64_t>;13:02
@HeikoSbut I get13:02
@HeikoSMathematics.i:25: Error: Template 'COperatorFunction' undefined.13:02
lisitsynHeikoS: I think some include is missed13:02
@HeikoS%include <shogun/mathematics/logdet/OperatorFunction.h>13:03
@HeikoS#ifdef USE_FLOAT6413:03
@HeikoS    %template(RealOperatorFunction) COperatorFunction<float64_t>;13:03
@HeikoSsame error13:03
@HeikoSdo I need to include something else than the class?13:03
lisitsynHeikoS: Mathematics_includes.i too IIRC13:03
@HeikoSwell its in there13:04
lisitsynHeikoS: oohh I think you have to push lambday a bit to structurize things in logdet13:05
@HeikoSlisitsyn: really?13:05
lisitsynI think there should be logdet/solver13:05
@HeikoSso is it possible to have a non-templated child class inheriting from a templated one?13:05
lisitsynmay be like that13:06
@HeikoSlisitsyn: ok will do13:06
@HeikoSyeah the solvers should be in math itself13:06
-!- van51 [] has quit [Quit: Leaving.]13:06
lisitsynHeikoS: do you think that's the problem?13:06
@HeikoSoperator too13:06
@HeikoSlisitsyn: no idea really13:06
@HeikoSI dont do this swig stuff often13:06
@HeikoSso a problem13:06
lisitsynHeikoS: okay let me check it13:07
@HeikoSCOperatorFunction<float64_t>   ->  CRationalApproximation   -> CLogRationalApproximationCGM13:07
@HeikoSthis is an inheritence thing going on13:07
lisitsynthat looks okay13:07
lisitsynin features we have the same thing13:07
@HeikoSthe CRationalApproximation fixes the template upon inheritance13:07
lisitsynHeikoS: ahh didn't you forget13:10
@HeikoSwhere to put it?13:10
@HeikoSyep that did it :)13:10
lisitsynHeikoS: \o?13:10
lisitsynI am a pirate lol13:11
lisitsynthe favourite thing of wiking13:11
@HeikoS/home/heiko/Desktop/shogun/shogun/src/shogun/mathematics/logdet/OperatorFunction.h:141: Warning 320: Explicit template instantiation ignored.13:11
@HeikoSany idea about this one?13:11
lisitsynHeikoS: ouhm13:12
@HeikoStemplate class COperatorFunction<bool>;13:12
@HeikoStemplate class COperatorFunction<char>;13:12
@HeikoSits those guys13:12
lisitsynwait wait13:12
lisitsynwhy in h?13:12
lisitsynHeikoS: I think it is .cpp13:13
lisitsynsee dense features13:13
@HeikoSyep solved it13:17
@HeikoSlisitsyn: how chan I check for two #ifdef?13:22
@HeikoSjust type it twice?13:22
lisitsynHeikoS: #if defined(BLA) && defined(SHHH)13:22
@HeikoSlisitsyn: do you know whether USE_COMPLEX64 is defined?13:23
lisitsynHeikoS: no I don't know due to latest changes13:23
@HeikoSwiking: shogun just used my local installation for compiling rather than the local files :(14:05
@HeikoShad to delete them by hand14:05
sonne|workHeikoS: file a bug assign it to wiking and tag it cmake14:41
sonne|workHeikoS: I had this issue too IIRC14:41
lisitsynandreas had it14:41
sonne|workso everyone :D14:42
@HeikoSlisitsyn: still there?15:17
@HeikoSanother swig issue:15:18
@HeikoScc1plus: warning: unrecognized command line option "-Wno-c++11-narrowing" [enabled by default]15:18
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:108: error: 'CKernelRidgeRegression' is not a member of 'shogun'15:18
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:140: error: 'self' was not declared in this scope15:18
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:163: error: expected primary-expression before '*' token15:18
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:164: error: 'data' was not declared in this scope15:18
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:175: error: expression list treated as compound expression in initializer [-fpermissive]15:18
lisitsynHeikoS: ja15:18
@HeikoSI get loads of those15:18
@HeikoS/usr/include/python2.7/numpy/npy_deprecated_api.h:11:2: warning: #warning "Using deprecated NumPy API, disable it by #defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION"15:18
lisitsynHeikoS: can you paste da line?15:18
@HeikoSda line?15:18
@HeikoSfrom /home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:6027:15:19
lisitsynyeah am speaking like west side nigga15:19
@HeikoSand I gets 100s of those ya15:19
lisitsynHeikoS: this one15:19
@HeikoSgedit crashes haha15:20
lisitsynHeikoS: use vim or nano or whatever15:21
lisitsynthey'll handle that15:21
@HeikoSyes doing15:21
@HeikoSSWIGINTERN shogun::CRegressionLabels *shogun_CKernelRidgeRegression_apply__SWIG_0(shogun::CKernelRidgeRegression *self,shogun::CFeatures *data=NULL){15:22
@HeikoS            return self->apply_regression(data);15:22
@HeikoS/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/modshogunPYTHON_wrap.cxx:9726:108: error: 'CKernelRidgeRegression' is not a member of 'shogun'15:22
@HeikoSlisitsyn: this has to do with my changes to Mathematics.i though15:23
@HeikoSshould I paste the content maybe?15:23
lisitsynHeikoS: really?15:23
lisitsynso w/o your changes it works?15:23
@HeikoSlisitsyn: yes15:24
@HeikoSif you replace yours with those youll get the error15:24
@HeikoSprobably syntax or something stupud15:25
-!- pickle27 [] has joined #shogun15:26
lisitsynHeikoS: haven't catched it yet15:27
@HeikoSseems almost like forgotten {15:27
@HeikoSin namespace15:27
lisitsynHeikoS: I'd expect that15:28
lisitsynbut all the namespaces are enclosed15:28
lisitsynpickle27: hey there15:28
pickle27lisitsyn: hey15:31
@HeikoSlisitsyn: what happens if you compile that?15:32
lisitsynHeikoS: I haven't checked15:32
@iglesiasghey pickle27! I wrote just a very minor comments in the pull request15:33
@iglesiasgpickle27, just to avoid some possible doxygen warnings15:33
pickle27iglesiasg: I saw, I'll make those changes today15:34
@iglesiasgpickle27, good, thank you :)15:34
pickle27iglesiasg: made those changes15:58
@iglesiasgpickle27, cool. We let travis finish and merge then16:05
pickle27iglesiasg: okay cool, I'll be back on later on!16:12
@iglesiasgpickle27, all right, see you!16:12
-!- pickle27 [] has left #shogun []16:12
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:8181:8528:e5a1:bac7] has quit [Quit: Ex-Chat]16:15
-!- foulwall [~zhengyang@] has joined #shogun16:17
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:48d3:5035:94d7:36e3] has joined #shogun16:18
-!- mode/#shogun [+o iglesiasg] by ChanServ16:18
@HeikoSiglesiasg: there?16:30
@iglesiasgHeikoS, yeah16:30
@HeikoSiglesiasg: do you know whether this voting system you used allows multiple votes per person?16:30
@HeikoSI would like 2 :)16:30
@iglesiasgHeikoS, I am not sure16:31
@iglesiasgHeikoS, I could check it in some minutes if you like16:31
@HeikoSand people would have to vote for two different names16:31
@HeikoSwould be nice, currently writing an email to set this up16:31
@iglesiasgHeikoS, all right, I am taking care of another thing right now but will check pretty soon!16:32
-!- lisitsyn [] has quit [Quit: Leaving.]16:41
@iglesiasgHeikoS, I found this16:52
@iglesiasgHeikoS, The voter will receive an email containing a unique vote ticket and a series of links to vote and display the results of the survey. Naturally a voter can only vote once.16:52
@iglesiasgHeikoS, but there is something about multiple choices16:53
-!- sonne|osx [~sonne@] has joined #shogun17:01
@iglesiasgHeikoS, OK, so it seems it works fine17:01
@iglesiasgHeikoS, I just did a test survey and invited you too.17:02
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:48d3:5035:94d7:36e3] has quit [Quit: Ex-Chat]17:07
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:a8ba:617a:d7c8:e097] has joined #shogun17:08
-!- mode/#shogun [+o iglesiasg] by ChanServ17:08
-!- sonne|osx [~sonne@] has quit [Quit: sonne|osx]17:15
-!- shogun-notifier- [] has joined #shogun17:15
shogun-notifier-shogun: Soeren Sonnenburg :develop * 98a36cc / INSTALL:
shogun-notifier-shogun: update install for cmake17:15
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:a8ba:617a:d7c8:e097] has quit [Quit: Ex-Chat]17:19
-!- van51 [] has joined #shogun17:19
shogun-buildbotbuild #1561 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>17:24
shogun-notifier-shogun: Heiko Strathmann :develop * 29f75f2 / src/shogun/mathematics/logdet/ (8 files):
shogun-notifier-shogun: move external type definitions to new cpp files17:29
shogun-notifier-shogun: Heiko Strathmann :develop * 495f4b9 / src/shogun/mathematics/logdet/ (4 files):
shogun-notifier-shogun: put definitions into namespace17:29
shogun-notifier-shogun: Heiko Strathmann :develop * 82f9a91 / src/shogun/mathematics/logdet/ (8 files):
shogun-notifier-shogun: Merge pull request #1494 from karlnapf/develop17:29
shogun-notifier-shogun: move external type definitions to new cpp files17:29
shogun-buildbotbuild #1928 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>17:34
shogun-buildbotbuild #1929 of deb1 - libshogun is complete: Success [build successful]  Build details are at
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:8ddd:e10b:a80:45e7] has joined #shogun17:36
-!- mode/#shogun [+o iglesiasg] by ChanServ17:36
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:8ddd:e10b:a80:45e7] has quit [Client Quit]17:36
shogun-buildbotbuild #1562 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Heiko Strathmann <>17:40
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:4582:9b04:f3dc:4e3a] has joined #shogun17:45
-!- mode/#shogun [+o iglesiasg] by ChanServ17:46
@iglesiasgHeikoS, did you see the poll?17:46
@HeikoSiglesiasg: hi17:46
@HeikoSyes I did17:46
@HeikoSthanks for that17:46
@HeikoSiglesiasg: so I will ask you to do two polls soon17:46
@HeikoSone accept/reject the shogun satzung17:46
@HeikoSand one with the names of all members of the foundation17:47
@HeikoSwill send details then17:47
@iglesiasgHeikoS, ok17:48
@HeikoSiglesiasg: can you currently compile shogun?17:49
@HeikoSI got this weird error compi?ling a unit test17:49
@iglesiasgHeikoS, latest develop?17:49
@iglesiasgHeikoS, let me try17:50
@HeikoSProbingSampler unit test17:50
@iglesiasgHeikoS, mm I didn't get any error here17:56
@iglesiasglet me see if I can run that test17:57
@HeikoSiglesiasg: do you have eigen installed or bundled by cmake?17:58
@iglesiasgHeikoS, funny thing is that ./shogun-unit-test --gtest_filter=Probing*17:59
@iglesiasgdoesn't run any test17:59
@HeikoSiglesiasg: ah you dont have colpack installed17:59
@iglesiasgHeikoS, aah ok. It must be that yeah17:59
@HeikoSiglesiasg: could you install and try?17:59
-!- sonne|osx [] has joined #shogun18:00
@iglesiasgHeikoS, about Eigen I bundle it, the version in my system is older than the required18:00
@iglesiasgHeikoS, apt-get for colpack?18:00
shogun-buildbotbuild #1025 of rpm1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Soeren Sonnenburg <>18:00
@HeikoSno apt-get :)18:00
@HeikoSonly if no hassle for you18:00
@iglesiasgHeikoS, is it installed in travis?18:00
@HeikoSiglesiasg: I dont know, think not18:00
@iglesiasgHeikoS, how do I specify for Shogun where to find colpack?18:01
@iglesiasgdo we have finpack for it in cmake?18:01
@HeikoSiglesiasg: shogun detects automagically18:01
@HeikoSyou just need a local installation18:01
@iglesiasgHeikoS, I am building colpack at the moment18:02
@iglesiasgHeikoS, I will need to run cmake in Shogun afterwards, right?18:02
shogun-buildbotbuild #1346 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>18:03
shogun-buildbotbuild #62 of osx1 - libshogun is complete: Success [build successful]  Build details are at
@iglesiasgHeikoS, compiling again :)18:06
@iglesiasg#define HAVE_COLPACK 1 in my config.h18:07
shogun-notifier-shogun: Kevin :develop * 1a88d4c / src/shogun/ (18 files):
shogun-notifier-shogun: added more setter and getters to ICA base class and cleaned up the classes18:08
shogun-notifier-shogun: Kevin :develop * df3e8c5 / src/shogun/converter/ica/ (6 files):
shogun-notifier-shogun: added some descriptions for doxygen in ICA classes18:08
shogun-notifier-shogun: Fernando Iglesias :develop * e7cce11 / src/shogun/ (18 files):
shogun-notifier-shogun: Merge pull request #1490 from pickle27/develop18:08
shogun-notifier-shogun: added more setter and getters to ICA base class and cleaned up the class...18:08
@iglesiasgHeikoS, ping18:17
shogun-buildbotbuild #1647 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at  blamelist: Heiko Strathmann <>18:17
@HeikoSiglesiasg: pong18:17
@iglesiasgHeikoS, error in here18:17
@HeikoSiglesiasg: uh could you remind me how to tell python that I installed shogun locally?18:17
@HeikoSiglesiasg: ah thats good18:17
@iglesiasgHeikoS, an error that is good mmm18:18
@HeikoSiglesiasg: yes since that means its not weird stuff18:18
@HeikoSiglesiasg: so I export LD_LIBRARY_PATH, PYTHONPATH, what else=?18:18
@iglesiasgjust kidding :)18:18
@iglesiasgHeikoS, something related to these python directories18:18
@iglesiasgI don't remember their name mmm18:18
shogun-buildbotbuild #1563 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Kevin <>18:19
@iglesiasgHeikoS, dist-packages18:20
shogun-buildbotbuild #63 of osx1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Kevin <>18:20
@iglesiasgHeikoS, but what is it exactly you are trying to do? Tell python where your shogun because you cannot install with sudo?18:21
-!- van51 [] has quit [Quit: Leaving.]18:21
@HeikoSiglesiasg: ah I got all weird errors18:21
@HeikoSinstalling with sudo again18:21
@HeikoSiglesiasg: not its because shogun gets confused with includes when there is a local install18:21
@iglesiasgHeikoS, aham18:21
@HeikoSI had compile errors caused by installed headers being different to the source ones18:21
@iglesiasgHeikoS, there is one variable inside python that has to be modified IIRC18:22
shogun-buildbotbuild #64 of osx1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Fernando Iglesias <>, Kevin <>18:23
-!- pickle27 [] has joined #shogun18:24
shogun-buildbotbuild #1564 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Fernando Iglesias <>, Kevin <>18:25
shogun-buildbotbuild #1026 of rpm1 - libshogun is complete: Success [build successful]  Build details are at
@HeikoSiglesiasg: still there?18:41
@iglesiasgHeikoS, indeed18:41
@HeikoSiglesiasg: man I get annoyed18:41
@HeikoSfrom modshogun import18:42
@HeikoSImportError: cannot open shared object file: No such file or directory18:42
@HeikoSwith anything I do18:42
@HeikoSI compiled and installed to /usr/local18:42
@iglesiasgHeikoS, let me check one thing18:42
@HeikoSah the file is not there18:42
@HeikoSbut why18:43
@HeikoScmake tells me the file is up to date18:43
@HeikoSwhile its not there18:43
@HeikoSthat is weird18:43
@iglesiasgHeikoS, does your PYTHONPATH include the dist-packages directory inside your Shogun installation?18:44
@HeikoSiglesiasg: yes its in /usr/local18:44
@HeikoSbut the is not there18:45
@HeikoScmake doesnt install it18:45
@HeikoS-- Up-to-date: /use/local/lib/
@HeikoS-- Up-to-date: /use/local/lib/
@HeikoS-- Up-to-date: /use/local/lib/libshogun.so18:45
@HeikoSman I am stupid :D18:45
@HeikoSall this hacking, its not made for me18:45
@HeikoS"use/local ....18:45
@iglesiasgsubtle thing18:46
@iglesiasgI didn't notice either18:46
shogun-buildbotbuild #1347 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Heiko Strathmann <>18:46
@HeikoSman I get those problems all day18:46
@HeikoSsorry for complaining loudly here ;)18:46
@iglesiasgwe are here for this :)18:47
-!- lambday [67157e4c@gateway/web/freenode/ip.] has joined #shogun18:58
lambdayHeikoS: hi18:58
lambdaywhat problems did we face for swig typemaps?18:59
@HeikoSlambday: some minor things, see my PR18:59
@HeikoSlambday: I am just to code up a small python example18:59
@HeikoSI wrote interface files18:59
@HeikoSwant to test it :)18:59
lambdayHeikoS: awesome! :D19:00
@HeikoSlambday: yep lets see :)19:00
@HeikoSinterfaces are tricky19:00
@HeikoSlambday: we definitely need to provide some std cases so that people dont have to instanciate all these classes19:00
@HeikoSok I will try it out now19:00
lambdayHeikoS: so, for the abstract base classes, if we tell the compiler to create with specific types in the header, it creates problem?19:00
@HeikoSlambday: btw the unit test for the probing sampler doesnt work19:00
@HeikoSlambday: no we need a convienience constructor which we just give a matrix which sets up default configurations19:01
lambdayHeikoS: doesn't work? :-/19:01
@HeikoSwithout any classes19:01
@HeikoSlambday: compile error19:01
@HeikoSProbingSampler test19:01
@HeikoSsome eigen static compile error19:01
@HeikoSnot detected by travis19:01
@HeikoSsince no colpack on there19:01
@HeikoSits the eigen version that comes with cmake bundle19:01
@HeikoSand my local one too19:01
@HeikoSiglesiasg: could reproduce it19:02
lambdaybut the guarded the whole test with #ifdef #endif19:02
lambdayfor colpack :-/19:02
lambdaynot sure HeikoS last night I tested on the machine that sonne|osx gave access to, and it worked.. I'll check into it19:02
lambdayHeikoS: could you please try installing colpack and see what it says? :(19:03
@HeikoSlambday: I have colpaclk installed19:03
@HeikoSits not a colpack problem19:03
@HeikoSthe guards work fine19:03
@HeikoSits an eigen problem19:03
@HeikoSlambday: let me finish this and then tell you more details19:04
@HeikoSlambday: dont worry :)19:04
lambdayHeikoS: alright :)19:04
lambdayso, I should add new tests as libshogun tests in undocumented, right? or as unit-tests....(I wanna add tests for the ozone data)'19:05
lambdaylibshogun *examples*19:05
-!- van51 [] has joined #shogun19:06
shogun-buildbotbuild #1648 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at
@HeikoSlambday: what?19:08
@HeikoSlambday: which class computes the number of shifts again?19:08
lambdayHeikoS: RationalApproximation iirc19:08
lambdaychecking again19:08
lambdayoh no, number of shifts is a user input19:09
@HeikoSnum shifts are a parameter of CLogRational ...19:09
lambdaywait :-/19:09
@HeikoSlambday: is there a method to compute the number?19:09
@HeikoSI remember the eliptic curve functions19:09
lambdayyes its in one of the unit-tests19:09
@HeikoSah no there was this error bound thing19:09
lambdayyes that has to be user input19:09
shogun-buildbotbuild #1649 of deb3 - modular_interfaces is complete: Failure [failed install test python_modular]  Build details are at  blamelist: Fernando Iglesias <>, Kevin <>19:10
@HeikoSshould be possible to automate this (optional)19:10
@HeikoSor in fact default should be automatic and user provides accuracy19:10
@HeikoSlambday: some other things: could you structure the logdet classes a bit19:10
@HeikoSlambday: in fact, most things can go into own folders19:11
@HeikoSlambday: such as operators/ solvers/ etc19:11
lambdayHeikoS: yes they should..19:11
@HeikoSsubfolders in mathematics19:11
lambdayHeikoS: alright, I will do that...19:11
lambdayyes that would be better19:11
@HeikoSlambday: but wait until I submitted the python example19:11
@HeikoSotherwise interfaces will break19:11
@HeikoSthis way we can detect that19:11
@HeikoSlambday: just realising how much work you did there, all these classes ... :D19:12
lambdaybut I am a bit concerned about the performance19:12
lambdayin many places I could have used const type& but I used pass-by-val19:12
@HeikoSlambday: for what type of arguments?19:13
lambdayand I am most concerned about the performance of cg solvers..19:13
@HeikoSwell see19:13
lambdaySGVector<T> mostly19:13
@HeikoSlambday: nice that you have all these checks19:13
@HeikoSno segfaults if something is NULL19:13
lambdaythose are the ones I am worried about19:13
@HeikoSlambday: should be fast19:13
@HeikoSbut if you can change a few, thats useful19:13
@HeikoSbut later19:13
@HeikoSfirst things have to work19:13
lambdayHeikoS: so, I should add the ozone example in example/libshogun/undocumented/., right?19:14
@HeikoS ** On entry to DSTEMR parameter number  8 had an illegal value19:15
lambdaywait checking19:16
@HeikoSlambday:  let me push the interfaces19:16
lambdayHeikoS: you meant this one?'19:17
@HeikoShere is how this looks19:18
@iglesiasg hushell, what is the next thing to do in the FactorGraph model??19:18
@HeikoSlambday: whats this error, have you seen that?19:19
lambdayHeikoS: nope!19:19
@HeikoSlambday: I bet thats colpack19:19
lambday8th param in DSTEMR?19:19
@HeikoSthats what it prints19:19
lambdaythat cannot be colpack, that's lapack!19:20
@HeikoSwhat? :)19:20
lambdayand where I used it for, it just takes the length of an SGVector19:20
lambdayhow can that be illegal when I already initialized it with a std::vector<T>::size()19:20
@HeikoSlambday: why are you using lapack there?19:21
@HeikoSmax eigenvalue?19:21
@HeikoSwhy lapack?19:21
lambdayHeikoS: that's for the triangular eigensolver19:21
lambdayyes yes19:21
@HeikoSeigen cannot do that?19:21
lambdayHeikoS: because eigen3 didn't have anything for that19:21
@HeikoSwhat matrix is that again?19:21
lambdayyes it cannot :(19:21
lambdayany psd matrix19:21
@HeikoShow large?19:22
pickle27wiking: around?19:22
lambdayfor lanczos, we do CG iterations, and we put two params as main diagonal and subdiagonal as another matrix (lanczos T-matrix) which is a tridiagonal matrix19:22
@HeikoSlambday: if this is a dense matrix, how does lapack do it? iteratively or ?19:22
lambdayHeikoS: depends on the number of iteration19:22
lambdayHeikoS: no idea! I took that as a blackbox19:23
@HeikoSi remember19:23
@HeikoSlambday: since for dense matrices that are not too large, computing all eigenvalues is not that much more expensive that just one19:23
@HeikoSso could also compute all and take the maximum, but this depends on matrix size19:23
lambdaybut this worked awesome with all the machines I tried it in19:24
pickle27wiking: nvm think I found my mistake, was playing with cmake!19:24
lambdayHeikoS: where does it give error exactly? for modular interface?19:24
lambdaydstemr is awfully accurate19:24
lambdayHeikoS: yes I thought of that, but when we can manage with two vectors of size 1000xsizeof(double), why should we use 1000x1000xsizeof(double) matrix, right?19:25
@HeikoSlambday: yeah maybe you are right19:26
@HeikoSIll try a bit further19:26
lambday1000=number of iteration (and therefore the default dim for Lanczos T-matrix)19:26
lambdayHeikoS: if it gives error, we need to test it more19:26
lambdaymay be something I missed19:26
lambdaythe typemaps looks awesome, although I didn't fully get how it works19:28
lambdayand yeah, since we don't need internal complex linear operators, we don't need to expose them to modular typemaps19:28
lambdayjust the directlinearsolvercomplex19:29
shogun-buildbotbuild #1348 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Fernando Iglesias <>, Kevin <>19:29
lambdayI was watching these video lecture series by Alexander Stepanov, man he is really against virtual methods! :-/19:31
shogun-buildbotbuild #1646 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at  blamelist: Soeren Sonnenburg <>19:33
@HeikoSlambday: why?19:36
lambdayHeikoS: he's convinced that its slow and there is better way to do it (for that I need to watch the whole series)19:37
lambdayin fact he says that he hasn't seen a really good example where OOP can *really* be a lifesaver19:37
@HeikoSlambday: and the alternative is?19:38
lambdayI have no idea :(19:38
lambdayI gotta watch the whole series and then may be I can say something.. but is it slow?19:38
lambdayI mean, I am not so sure how does compile treats virtual functions19:39
@HeikoSlambday: these are issues for the compiler architects :)19:39
@HeikoS-c:16: RuntimeWarning: [WARN] In file /home/heiko/Desktop/shogun/shogun/src/shogun/mathematics/logdet/LanczosEigenSolver.cpp line 145: Some error occured while computing eigenvalues!19:39
lambdaybut to me, its obviously the cleaner way to do it - specially with the situation like this that we have for log-det19:39
@HeikoSmatrix is identity 10x1019:39
@HeikoSlambday: yeah dont worry about virtual methods19:40
lambdayHeikoS: it says it doesn't converged??19:40
@HeikoSlambday: coding time also counts19:40
@HeikoSlambday: this is the error I get19:40
@HeikoSwarning rather19:40
@HeikoSI still get a reasonable result from the log-det sampler19:40
lambdayHeikoS: that is probably because I set the default convergance cond for all CG based solver to be 1e-619:40
@HeikoSbut thats identity matrix19:41
lambdayon residual'19:41
lambdaydoesn't the warning say anything else?19:41
lambdaywait checking19:41
lambdayif (info==0)19:42
lambdaydo you have cblas installed?19:42
lambdayit seems the dstemr you have isn't right... for some reason!19:43
-!- HeikoS [] has quit [Remote host closed the connection]19:43
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:4582:9b04:f3dc:4e3a] has quit [Quit: Ex-Chat]19:43
-!- HeikoS [] has joined #shogun19:44
-!- mode/#shogun [+o HeikoS] by ChanServ19:44
@HeikoSlambday: sorry computer crash19:45
lambdayHeikoS: np :)19:45
lambdayHeikoS: I am worried that the lapack you have does the right thing!19:45
lambdaybecause info should NOT be non-zero!19:46
@HeikoSlambday: we will find loads of those19:46
lambdaythat's helpful actually19:46
@HeikoSlambday: indeed19:48
lambdayHeikoS: where do you have
@HeikoSthere are more problems in fact19:48
lambdayHeikoS: more? :( like?19:48
lambdayand not /usr/lib64/atlas/
lambdayI am not the guy to ask but sergey can may be say something!19:49
lambdayand if it turns out that it depends on atlas so we should guard it with proper #ifdefs19:50
@HeikoSlambday: do you have latest source compiled?19:52
@HeikoSlambday: with python bindings?19:52
lambdaywait I am testing19:52
@HeikoSthis is a little script that gives wrong results, maybe I have done something wrong in setting the thing up19:54
@HeikoSlambday: but maybe first add the automatic num_shifts19:54
lambdayHeikoS: so, inside the RationalApproximation class19:55
@HeikoSfirst thing to do is to make a very simply example work19:55
@HeikoSlambday: and the shifts are crucial for that19:55
lambdayHeikoS: in one unit-test, I did that19:56
@HeikoSlambday: yeah I know, then you can use that code19:56
lambdayHeikoS: so, I should add a python modular example that does the same thing, right?19:57
lambday(btw this is the example,
@HeikoSlambday: nono, lets wait with the examples19:57
@HeikoSadd a method to the corresponding class19:58
@HeikoSand the user specified num shifts should be optional19:58
lambdayHeikoS: alright...19:58
@HeikoSif you make a method that returns needed number of shifts or so19:58
@HeikoSthen user can even check that and the decide what to do19:58
@HeikoSlambday: next thing is to make my gist work19:58
@HeikoSit returns the wrong result19:58
@HeikoSlambday: there should also be a unit test for this case19:59
@HeikoSthat is in the python example19:59
lambdayHeikoS: I will test...19:59
lambdayHeikoS: do you think it gives wrong result because of the lapack eigensolver thing?20:00
@HeikoSlambday:  no thats another problem20:00
@HeikoSlambday: this happened when I passed a all zeros matrix20:00
@HeikoSor all ones20:00
@HeikoSlambday: which is ok not to work, but there should be a check for these cases20:00
lambdayHeikoS: aaaarghhh! I think the shifts computation method doesn't work then eig_max/eig_min=120:01
@HeikoSlambday: so maybe add on your list "unit test for stupid borderline cases" :)20:01
lambdayheheh yes I will20:01
@HeikoSlambday: yeah, this should be catched and a proper message should be output20:01
@HeikoSsince users as me are doing stupid things all the time :D20:01
lambdayyes point noted20:01
lambdayHeikoS: could you please try with a different diagonal matrix and test if that at least works?20:02
@HeikoSlambday: Ill update the gist20:02
@HeikoSthis one uses random positive numbers on the diagonal20:03
lambdaythis too gives poor result?20:03
lambdayany warnings on the way?20:04
lambdayfor non-convergance?20:04
@HeikoSwrong result20:04
@HeikoSno warnings/errors20:04
@HeikoSoh CStatistics also give wrong results for a bit larger matrices20:05
lambdaythe direct one?20:05
lambdayHeikoS: upto rational approximation of f(A)*v, everything is really accurate20:05
lambdayafter that, things lose their shines :(20:06
@HeikoSlambday: well get there, maybe I did something wrong in the script20:06
@HeikoShave a look20:06
@HeikoSfor matrix size 10, CStatistics and sum(log(diag)) is the same20:06
@HeikoSbut the log-det estimator is off20:07
lambdayHeikoS: logdetestimator does the mean for you, right?20:07
@HeikoSlambday: but the shifts are set arbritary, this is the first thing to do20:07
@HeikoSlambday: yes20:07
lambdayyeah I should change that20:07
lambdayHeikoS: will do it by tomorrow20:07
-!- van51 [] has quit [Read error: Connection reset by peer]20:07
lambdayHeikoS: how much does it deviate?20:08
@HeikoSlambday: there is something wrong its completely off20:08
@HeikoSnot just inaccurate20:08
lambdayHeikoS: I can test this same thing with cpp, right?20:08
@HeikoSlambday: yes, best would to make unit tests from it20:09
@HeikoSlambday: but again, shifts might cause this, currently playing with them20:09
lambdayHeikoS: safe is 20+ I think20:09
lambdaybut it should not be *completely* off20:09
-!- van51 [] has joined #shogun20:10
lambdayHeikoS: I noticed few such cases with gaussian samplers... where the solvers just refused to converged and we got awfully off results20:10
@HeikoSoh, speaking of it20:12
@HeikoSthe normal sampler produces something even more weird20:12
@HeikoSwith the probing sapler there is also something wrong20:12
@HeikoSit is way to fast with it20:12
@HeikoSeven for 10000 samples20:13
@HeikoSwhereaus the gaussian sampler takes a while20:13
lambdaythat's weird! probing sampler gives #colors sampler per log-det-sample where normal samplers give 1 sample vector per log-det sample20:15
lambdayboth relies on randgen20:15
lambdayargh I should add more and more tests.. HeikoS you will have a bunch of tests by tomorrow20:16
@HeikoSlambday: it will be a very good idea to create more tests around this20:16
@HeikoSlambday: its also very good that you tested every module seperately20:16
@HeikoSlambday: now we just need to make sure things work well together20:16
lambdayHeikoS: that's why its weird that we get these errors :(20:16
@HeikoSlambday: so if you turn the pyton example into a unit tests that works, thats already a lot20:17
lambdayHeikoS: upto rational approximation of f(A)*v things just work perfect!20:17
lambdayHeikoS: yes I will do that20:17
@HeikoSlambday: then we can do multiple ones for all these different cases20:17
@HeikoSand then. ozone :)20:17
@HeikoSmy mcmc code is ready, just need the python function to compoute the log-det :)20:17
lambdayyes that's the final target20:17
@HeikoSwould be great if that worked by next week since I will present this at a workshop soon20:18
lambdayHeikoS: that's gonna be a part of shogun too?20:18
@HeikoSlambday: no not yet, maybe next gsoc20:18
lambdayHeikoS: whoa!20:18
lambdayHeikoS: will add these tests by tomorrow20:19
@HeikoSlambday: cool!20:19
lambdaytheoretically, this should not be the base20:19
@HeikoSI will be there tomorrow to help you on that20:19
lambdayso must be some technical silly mistake20:19
@HeikoSleaving to germany on the weekend though20:19
@HeikoSlambday: yeah I am sure it is, it always is20:19
@HeikoSits good that we run into those now, not later20:19
lambdayHeikoS: next week you guys are going to US, right? :)20:20
lambdayHeikoS: yes :-/20:20
@HeikoSno thats oct20:20
lambdayI starred your gists.. you will have the tests by tomorrow20:21
@HeikoSok I will go home now20:23
@HeikoSsee you tomorrow!20:23
lambdayHeikoS: see you, man! :)20:23
-!- HeikoS [] has quit [Quit: Leaving.]20:25
-!- lambday [67157e4c@gateway/web/freenode/ip.] has quit []20:28
-!- shogun-notifier- [] has quit [Quit: transmission timeout]21:08
-!- lisitsyn [] has joined #shogun21:13
-!- van51 [] has quit [Quit: Leaving.]21:20
@sonney2klisitsyn, do you recall what I said how expensive the flight is?21:26
lisitsynsonney2k: which flight?21:26
lisitsynsonney2k: your?21:27
lisitsyn66* something IIRC21:27
@sonney2klisitsyn, yeah no probably not21:27
@sonney2kI currently only find flights for 950EUR21:28
lisitsynthat's in log right?21:28
@sonney2kif I want to be on the same plane like heiko21:28
lisitsynsonne|work: sonne|work823 with lufhansa over london21:28
@sonney2klooks like it got more expensive now21:29
lisitsynsonney2k: what if not on the same plane?21:29
@sonney2kahh no that was not taking the same flight back when I posted this21:29
@sonney2klisitsyn, the cheapest I can get is I guess around 70021:29
lisitsynsonney2k: the cheapest for me is 1K$ and the one that fits the time is 1.3K$21:30
@sonney2kwell if 2 stops and 20 hour flight it would be 68921:30
@sonney2kbut actually cheapest is now 81321:30
@sonney2kso I guess 950 is OK21:30
@sonney2kyes EUR21:30
lisitsynsonney2k: we have 1100 per person or?21:31
@sonney2klet me check how much I paid last time21:31
lisitsynI don't rembmer21:31
@sonney2klisitsyn, I think only 2200$ total21:31
@sonney2kwe can try to get a bit more21:31
@sonney2klast year chris was coming from US -> cheap21:32
-!- gsomix_ [~gsomix@] has joined #shogun21:32
lisitsynsonney2k: whereto do they transfer the money?21:32
lisitsynthe PO thing?21:33
@sonney2kactually last year I paid 87621:33
@sonney2klisitsyn, well in the end to me21:33
lisitsynand you to transferred to chris?21:33
@sonney2klisitsyn, no idea how to get it to you21:33
@sonney2klisitsyn, yes21:33
lisitsynthat's a real problem21:33
lisitsynsonney2k: it can be paypal21:34
@sonney2klisitsyn, well I could go to an ATM in the us and give it to you in cash21:34
lisitsynif it works for you21:34
lisitsynyeah that's the easiest way21:34
lisitsynif it works for you21:34
@sonney2kok so I will book this expensive bitch21:34
@sonney2knext year we have to book more early21:35
@sonney2kI think we could have saved 300 EUR in total21:35
@sonney2kwhich is about what the server costs :/21:35
-!- gsomix__ [~gsomix@] has quit [Ping timeout: 245 seconds]21:36
lisitsynsonney2k: but we can't spend the money other way right?21:37
@sonney2klisitsyn, what do you mean?21:44
lisitsynsonney2k: we can use it only for flight tickets I mean21:44
@sonney2klisitsyn, yes21:48
@sonney2kwoah they added 17 EUR for paying with credit card and another 17 for issuing ticket22:02
lisitsynsonney2k: hah that's very sowjetische22:04
lisitsynsonney2k: why credit card is additional cost? would they instead like to see you in person and have your cash? :D22:05
-!- HeikoS [~androirc@] has joined #shogun23:03
-!- mode/#shogun [+o HeikoS] by ChanServ23:03
-!- iglesiasg [] has joined #shogun23:32
-!- mode/#shogun [+o iglesiasg] by ChanServ23:32
@iglesiasggood evening comrades23:32
@iglesiasgwe should find out what's happening with Kyle's machine23:34
pickle27yeah see if he can paste for us23:35
@iglesiasgit is because C++11 I think23:37
@sonney2kI guess so too23:38
@sonney2khe has been suffering for a bit too long :D23:38
@sonney2kactually :/23:38
@iglesiasgsonney2k, I just got an email to accept your answer in the mailing list23:38
@sonney2kiglesiasg, nah23:38
@sonney2kI switched my main machine23:38
@sonney2know all settings are gone23:38
@sonney2kI am finally back to work on a desktop23:38
@sonney2knotebooks suck :D23:39
@iglesiasgsonney2k, haha23:39
@iglesiasgpowerful machine?23:39
@sonney2kbetter than our new buildbot23:40
@sonney2kand it was idling here23:41
@iglesiasgwhy can it be that C++11 is not been recognized?23:41
@iglesiasgis there anyone around using gcc 4.4.7?23:42
@sonney2kiglesiasg, I think our old buildbot23:43
@sonney2k4.4.5 even23:44
@iglesiasgsonney2k, this one
@sonney2kone of them - I think slave323:45
@sonney2kyes slave323:45
@sonney2kthe problem is that c++11 features got slowly introduced by compilers23:46
@sonney2kso they did support only a bit of it at some point23:46
@iglesiasgsonney2k, but the thing I find weird is that I don't see these lines in Kyle's log23:46
@iglesiasg-- Performing Test _HAS_CXX11_FLAG23:46
@iglesiasg-- Performing Test _HAS_CXX11_FLAG - Failed23:46
@iglesiasg-- Performing Test _HAS_CXX0X_FLAG23:46
@iglesiasg-- Performing Test _HAS_CXX0X_FLAG - Success23:46
@iglesiasgmaybe he didn't paste them though23:47
@sonney2kwell CXX0X found - maybe someone used a CXX11 feature under the CXX0X brand23:47
@iglesiasgsonney2k, where do you see CXX0X found in Kyle's log?23:48
@iglesiasgwhat I pasted above is from the buildbot23:48
@sonney2k-- Performing Test _HAS_CXX0X_FLAG - Success23:51
@sonney2kiglesiasg, ok lets see what he says23:53
--- Log closed Fri Aug 30 00:00:37 2013