Open in new window / Try shogun cloud
--- Log opened Thu Nov 24 00:00:13 2011
--- Day changed Thu Nov 24 2011
@sonney2kfalse alarm?00:00
@sonney2kI mean how is it possible to have an invalid read of size one in this line00:00
blackburnsonney2k: ary/00:01
@sonney2kary is a ptr00:01
blackburnhow big is void pointer?00:01
@sonney2k8 bytes00:01
@sonney2kor 400:01
@sonney2kdepending on 64/32bit system00:01
blackburnsonney2k: then I don't know00:02
blackburnsonney2k: don't you get segfault with that script?00:02
@sonney2kdidn't try00:04
blackburncould you please?00:05
blackburnI'm currently recompiling and will test again00:05
blackburnsonney2k: don't you think testing framework is something of interest for gsoc project?00:07
@sonney2kyes agreed - but we need a concrete plan for that00:08
@sonney2kand it requires heikos migrations00:08
blackburnsonney2k: we have time to think about00:08
blackburnwow too late already00:13
blackburnsonney2k: yes I got segfault again00:17
blackburnwhat a non-sense00:17
@sonney2kthe other thing I'd like to see is a cool tutorial with applications00:17
@sonney2kblackburn, you should compile with optimizations disabled00:17
blackburnsonney2k: already00:17
@sonney2kand then the usual stuff00:18
@sonney2kdecision trees00:18
@sonney2kboosting etc00:18
@sonney2ksome new ML algos :)00:18
@sonney2kanyway I am tired00:19
@sonney2khave a nice sleep00:19
blackburnhave a nice sleep too00:19
blackburnsee you00:19
blackburnLabels with same numpy array leads to already freed memory error00:21
blackburnthat's crazy a lot00:21
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]00:28
-!- emrecelikten [~emre@] has joined #shogun05:28
emreceliktenHello there! Apparently Gaussian HMMs aren't supported in Shogun. Is there anybody working on it?05:29
-!- emrecelikten [~emre@] has left #shogun []06:59
-!- emrecelikten [~emre@] has joined #shogun06:59
sonne|workemrecelikten: noone is working on that09:08
sonne|workso you could be first :D09:08
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun14:13
blackburnsonne|work: forgot to say14:13
blackburne.g. Labels([1,2,3])14:14
blackburnwill lead to problems14:14
blackburnand we should do Labels(numpy.array([1,2,3]))14:14
sonne|workso the numpy one works the other not?14:14
blackburnI have too weak ideas what is going wrong but I know it is14:15
blackburnsonne|work: I managed to run this python example yesterday14:15
sonne|workgreat that you figured taht out14:16
blackburnsonne|work: if you have shogun available could you please run this:14:17
blackburnx = [1,2,3]14:17
blackburnfor i in range(10):14:17
blackburn  l = Labels(x)14:17
blackburnif it fails I would get more sure about it14:17
sonne|workI recall that we had a similar report some time back but no one could reproduce the problem14:17
blackburnI shared the folder with that script to madeleine, no report yet14:23
blackburnbut I hopt it will go smooth this time14:23
-!- blackburn_ [5bdfb203@gateway/web/freenode/ip.] has joined #shogun15:41
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Ping timeout: 265 seconds]15:43
-!- policecar [~policecar@] has joined #shogun15:44
-!- policecar [~policecar@] has quit [Quit: policecar]15:51
-!- blackburn_ [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]16:01
-!- policecar [~policecar@] has joined #shogun16:04
policecarhey, sorry to bother you(s) again but i can't seem to get PCA to work with sparse matrices... i currently run the code below (python) and can't get rid of17:18
policecarassertion features->get_feature_class()==C_SIMPLE failed in file preprocessor/PCA.cpp line 7117:18
policecarfrom shogun.Features import SparseRealFeatures, SparseIntFeatures17:18
policecar        from shogun.Preprocessor import PCA17:18
policecar        feats = SparseRealFeatures( data.astype(float) )17:18
policecar                17:18
policecar        preprocessor = PCA()17:18
policecar        preprocessor.init( feats )17:18
policecar        preprocessor.apply_to_sparse_feature_matrix( feats )17:18
policecarwhere data is a csc_matrix17:18
policecarcould you gimme any hints as to what i'm doing wrong?17:18
sonne|workit doesn't work with sparse matrices - so you need to convert it to dense features and then apply PCA17:21
policecarhehe, k, thx17:21
-!- policecar [~policecar@] has quit [Quit: policecar]17:29
-!- blackburn [~blackburn@] has joined #shogun17:30
blackburnsonney2k: my simple labels test does not reproduce the error anyhow19:38
@sonney2kblackburn, it does22:06
@sonney2kI can reproduce it here22:06
@sonney2kbut it is not reliable22:06
blackburnsonney2k: hmm ok22:06
@sonney2kthe issue occurs from time to time only22:06
@sonney2kif I use Labels([0.0]*1000) it happens most of the time22:07
blackburnsonney2k: I guess it is something of 'list' type nature22:07
@sonney2kbut I totally don't understand why it crashes sometimes and sometimes not22:07
@sonney2kyes but it is converted to narray22:08
blackburnsonney2k: is it?22:08
@sonney2kbut it crashes even before that(!!!)22:08
blackburnwell no idea then22:08
@sonney2kyes via PyArray_FromAny22:08
@sonney2kme neither22:08
blackburnno one should use lists to have a conversation with shogun :)22:09
blackburn1st commendment22:09
@sonney2kwe could simply disable that22:10
@sonney2klet me just do that22:10
blackburnsure, pretty unsafe22:10
@sonney2kI still don't understand how it could crash22:11
@sonney2k(I am in train home - network could fail at any time)22:12
blackburnyou really do shogun all the time :D22:12
@sonney2kstack overflow or so?22:18
@sonney2kargh need to leave train22:19
blackburnsonney2k: ok22:19
blackburnsonney2k: what stackoverflow?22:19
@sonney2kblackburn, I mean I cannot make sense of this valgrind error nor the crash23:32
@sonney2kand I don't understand why it still wants to accept lists23:33
@sonney2kmy only idea now is to explicitly reject lists23:33
blackburnwhy not23:34
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]23:51
--- Log closed Fri Nov 25 00:00:59 2011