--- Log opened Mon Jun 03 00:00:29 2013 | ||
van51 | sonney2k: I made a PR about the CCombinedDotFeatures | 00:01 |
---|---|---|
van51 | sonney2k: review it when you have time and let me know if we should run an example again | 00:02 |
-!- iglesiasg [~iglesiasg@213.143.50.52] has joined #shogun | 00:18 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 00:19 | |
-!- van51 [~van51@athedsl-320452.home.otenet.gr] has left #shogun ["PING 1370216909"] | 01:48 | |
-!- n4nd0 [~iglesiasg@213.143.50.52] has joined #shogun | 02:07 | |
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has quit [Ping timeout: 245 seconds] | 02:08 | |
-!- n4nd0 [~iglesiasg@213.143.50.52] has quit [Client Quit] | 02:09 | |
-!- iglesiasg [~iglesiasg@213.143.50.52] has quit [Ping timeout: 252 seconds] | 02:11 | |
-!- nube [~rho@49.244.100.165] has quit [Quit: Leaving.] | 03:38 | |
-!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has joined #shogun | 04:03 | |
-!- foulwall [~foulwall@2001:da8:215:503:cc30:1c53:ab9d:5f3c] has joined #shogun | 04:29 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 05:27 | |
-!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has quit [Ping timeout: 246 seconds] | 06:51 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Remote host closed the connection] | 07:20 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 07:43 | |
-!- gsomix [~gsomix@46.20.65.107] has quit [Quit: Leaving] | 08:21 | |
-!- sonne|work [~sonnenbu@sams-office-nat.tomtomgroup.com] has joined #shogun | 09:10 | |
sonne|work | hey foulwall! | 09:10 |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has joined #shogun | 09:23 | |
-!- gsomix [~Miranda@r206-10.smr.ru] has joined #shogun | 09:23 | |
gsomix | good morning! | 09:23 |
gsomix | sonne|work: "Our Work Is Never Over"? :) | 09:27 |
sonne|work | gsomix: what do you mean? | 09:27 |
gsomix | sonne|work: it's just question "are you working?" in sophisticated form. text from Daft Punk composition. nevermind. | 09:31 |
sonne|work | gsomix: yes indeed I am at work and even working ;) | 09:40 |
sonne|work | gsomix: btw your PR is not there yet right? | 09:40 |
sonne|work | gsomix: I was thinking that maybe we can speed up things a bit more even but lets get the initial version done first | 09:41 |
gsomix | sonne|work: there is. I have updated it at night. https://github.com/shogun-toolbox/shogun/pull/1130 | 09:42 |
sonne|work | gsomix: ohh please write a test for that | 09:44 |
sonne|work | btw I didn't realize that you submitted the PR... | 09:44 |
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has joined #shogun | 09:45 | |
gsomix | sonne|work: ok. this is unreadable again? :( | 09:45 |
gsomix | *is this | 09:48 |
sonne|work | gsomix: no no but we need a test | 09:49 |
sonne|work | gsomix: for both the circular buffer and the line reader | 09:49 |
gsomix | sonne|work: ok. which data set I shoudl use for testing? Human.dat is just bzipped. | 09:52 |
sonne|work | gsomix: just read the source code of the test with a very small buffer | 09:52 |
-!- gsomix [~Miranda@r206-10.smr.ru] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org] | 10:06 | |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has quit [Quit: Page closed] | 10:33 | |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has joined #shogun | 10:37 | |
-!- gsomix [~Miranda@r206-10.smr.ru] has joined #shogun | 10:44 | |
@wiking | vgorbati: ping | 10:52 |
@wiking | vgorbati: the problem with the rng does still exists i guess | 10:53 |
@wiking | right? | 10:53 |
sonne|work | wiking: what? | 10:55 |
-!- gsomix [~Miranda@r206-10.smr.ru] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org] | 10:55 | |
vgorbati | wiking: I thought so | 10:55 |
@wiking | sonne|work: there was some error with rng generation for one module | 10:55 |
@wiking | vgorbati: told me so | 10:55 |
@wiking | i mean across different OSes | 10:56 |
vgorbati | wiking: yes I did. but sonney|work convinced me, that it is not rng problem, since unit tests for rng on my machine were ok | 10:56 |
vgorbati | wiking: the output for one integration test was definitely different on different OSes, so I thought it was because of a rng. But it looks like the problem is in some other stuff | 10:58 |
-!- gsomix [~Miranda@r206-10.smr.ru] has joined #shogun | 10:58 | |
@wiking | vgorbati: hehehe that could be for a lot of other reasons... but then again it's a good guess to start with PRNG | 10:58 |
@wiking | the problem is that PRNG is implemented within shogun now | 10:59 |
@wiking | so it really should not | 10:59 |
@wiking | differ amongs different os/archs | 10:59 |
@wiking | gsomix: daft punk \o/ | 11:00 |
vgorbati | wiking: yes, that was the obvious reason to start with. The test is for dimension reduction method in tapkee, that involves random number generation, but it also involves other stuff, so some investigation is needed | 11:00 |
@wiking | vgorbati: lisityn will take care of it i hope :) | 11:01 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 11:22 | |
-!- mode/#shogun [+o lisitsyn] by ChanServ | 11:22 | |
sonne|work | wiking: speaking of which there he is | 11:25 |
sonne|work | lisitsyn: any news on the test bugfix (vgorbatis code) | 11:26 |
sonne|work | vgorbati: btw just do neural networks then - but make sure you check out torch{>=5} and lush and theano to see how others do it | 11:27 |
sonne|work | vgorbati: maybe you can even borrow some code :) | 11:27 |
vgorbati | sonne|work: yes, sure. | 11:29 |
@lisitsyn | sonne|work: argh no news sorry | 11:32 |
sonne|work | lisitsyn: btw how are your teeth? | 11:32 |
@lisitsyn | sonne|work: hmm quite ok why? | 11:33 |
@lisitsyn | sonne|work: ahh you mean these month-long dentist visits? | 11:33 |
sonne|work | yes your chill out sessions | 11:33 |
@lisitsyn | I should continue actually | 11:34 |
@lisitsyn | :D | 11:34 |
@lisitsyn | sonne|work: I accidentally got to pass some exams | 11:34 |
@lisitsyn | that's why I am losing this test thing | 11:34 |
sonne|work | lisitsyn: ohh nice I hope it was worth being lazy >:-> | 11:35 |
@lisitsyn | sonne|work: ohh I should kick you too then! :) | 11:36 |
@lisitsyn | sonne|work: what's with welcome mail for workshop? | 11:36 |
@lisitsyn | you told me you will fix it | 11:36 |
-!- foulwall [~foulwall@2001:da8:215:503:cc30:1c53:ab9d:5f3c] has quit [Remote host closed the connection] | 11:37 | |
sonne|work | ohh kick ass time | 11:37 |
gsomix | sonne|work: will be nice photo/video from workshop? | 11:39 |
@lisitsyn | german video you know | 11:40 |
gsomix | lisitsyn: ;) | 11:40 |
gsomix | this will be enough | 11:41 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has joined #shogun | 11:42 | |
@lisitsyn | sonne|work: I am thinking of some cloud showcase for the workshop | 11:42 |
@lisitsyn | we should discuss contents of our talks end of this month I think | 11:43 |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 11:43 | |
@lisitsyn | iglesiasg: how is it going? | 11:43 |
@iglesiasg | lisitsyn: hi! all fine, what about you? | 11:44 |
@lisitsyn | iglesiasg: fine too - is your stuff ready? | 11:47 |
sonne|work | stuff == thesis? | 11:48 |
@lisitsyn | yes sonne|work | 11:49 |
@iglesiasg | lisitsyn: yeah! waiting for my supervisor to give feedback | 11:49 |
@iglesiasg | lisitsyn: want to give a glance to it? | 11:50 |
@lisitsyn | iglesiasg: yes would be happy to! | 11:50 |
@iglesiasg | ok, give me a second | 11:50 |
@lisitsyn | iglesiasg: btw did you prepare something on your berlin trip? | 11:50 |
@lisitsyn | 1.5 months left! | 11:51 |
@iglesiasg | lisitsyn: I am on it | 11:51 |
@iglesiasg | lisitsyn: https://dl.dropboxusercontent.com/u/11020840/pfc.pdf | 11:51 |
@iglesiasg | normally these projects are rather long in here... | 11:52 |
@iglesiasg | for me that's way too many pages but it is actually short compared to some people present around here | 11:52 |
@lisitsyn | iglesiasg: your acknowledgements is empty! supervisor would be unhappy ;) | 11:52 |
@lisitsyn | el aprendizaje estructurado! | 11:53 |
@iglesiasg | haha | 11:53 |
@lisitsyn | iglesiasg: I like the layout pretty much | 11:53 |
@iglesiasg | lisitsyn: thanks! | 11:53 |
@iglesiasg | I cpied it from Patrick's thesis :P | 11:53 |
@lisitsyn | iglesiasg: is it standard or yours? | 11:53 |
@lisitsyn | ahh | 11:53 |
@lisitsyn | iglesiasg: are you free to play around with your layout? | 11:54 |
@iglesiasg | he had a git repo with a chapter from his thesis | 11:54 |
@iglesiasg | to some extent | 11:54 |
@iglesiasg | basically cover is fixed | 11:54 |
@lisitsyn | I would be punished for such layout here | 11:55 |
@lisitsyn | :D | 11:55 |
@iglesiasg | why? | 11:55 |
@lisitsyn | well they won't let me in | 11:55 |
@iglesiasg | I think it looks nice | 11:55 |
@lisitsyn | yes | 11:55 |
@lisitsyn | but they don't want nice looking | 11:55 |
@lisitsyn | they want GOST :) | 11:55 |
@lisitsyn | standard that is | 11:55 |
@iglesiasg | lisitsyn: have you checked some figures? | 11:55 |
@iglesiasg | tikz all the way! | 11:55 |
@lisitsyn | tikz? really? | 11:55 |
@iglesiasg | and some others with shogun | 11:56 |
@lisitsyn | tikz is such a pain | 11:56 |
@iglesiasg | see page 20 | 11:56 |
@iglesiasg | simple stuff but it looks nice I think | 11:56 |
@lisitsyn | yeah | 11:57 |
@lisitsyn | it just takes too much time for me | 11:57 |
@iglesiasg | and page 45 with Viterbi decoding | 11:57 |
@lisitsyn | to produce something reasonable | 11:57 |
@iglesiasg | Patrick has some really fancy stuff with tikz in his thesis | 11:58 |
@lisitsyn | iglesiasg: you may also put our paper here | 11:59 |
@lisitsyn | at least we can put it to arxiv | 11:59 |
@iglesiasg | lisitsyn: can I put a reference to sth not published yet?? | 11:59 |
@iglesiasg | ok | 12:00 |
@lisitsyn | yeah with some note | 12:00 |
@iglesiasg | it sounds nice :) | 12:00 |
@lisitsyn | I can submit it to arxiv right away | 12:00 |
@lisitsyn | sonne|work: am I right it is ok to submit a paper considered by jmlr to arxiv? | 12:00 |
sonne|work | lisitsyn: not sure but I think so | 12:01 |
@lisitsyn | "Authors may post their submitted manuscripts to open e-print archives, such as arXiv." | 12:03 |
@iglesiasg | lisitsyn: I can take care of it if you wish | 12:06 |
lambday | while checking switch over all ptypes, is it okay to keep those null (like SGObject) for features and stuffs? | 12:13 |
lambday | since we'll only need complex64_t for matrix and vector | 12:13 |
lambday | and a few math functions... | 12:13 |
@lisitsyn | iglesiasg: how much time do you have to change it a bit? | 12:13 |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 12:13 | |
@lisitsyn | iglesiasg: no I mean it could take a bit of time | 12:14 |
@lisitsyn | to publish it at arxiv | 12:14 |
@iglesiasg | lisitsyn: aham I see | 12:14 |
@lisitsyn | but yeah you are free to submit it too | 12:14 |
@iglesiasg | lisitsyn: I rather take care of cleaning all my branch first and fix a couple of things so I can merge my PR first | 12:15 |
@lisitsyn | I hereby grant you such permission ;) | 12:15 |
@iglesiasg | haha ok | 12:15 |
@lisitsyn | I think any of us is granted with such permission actually | 12:15 |
@lisitsyn | so if you want submit it right away | 12:15 |
@iglesiasg | lisitsyn: but do we have to change some stuff like layout or so? | 12:15 |
@lisitsyn | iglesiasg: yes I think so | 12:29 |
@lisitsyn | some arxiv speciifc stuff | 12:29 |
@iglesiasg | I see | 12:34 |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has quit [Quit: Page closed] | 13:21 | |
-!- HeikoS [~heiko@nat-160-98.internal.eduroam.ucl.ac.uk] has joined #shogun | 13:29 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 13:29 | |
lambday | HeikoS: hi | 13:30 |
@HeikoS | lambday: hi! | 13:30 |
@HeikoS | how is it going? | 13:30 |
lambday | HeikoS: how are you? | 13:30 |
@HeikoS | I am tired from the weekend ;) | 13:30 |
@HeikoS | how are you? | 13:30 |
lambday | I'm good... :) | 13:31 |
@HeikoS | how are the complex things going? | 13:31 |
lambday | for switch over all ptypes, I've kept it NULL just like SGObject in the class_list.cpp.py... because we don't need it for the features and stuff | 13:32 |
lambday | matrix, vector, sparse matrix, sparse vector - done | 13:32 |
lambday | math - done | 13:32 |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 13:32 | |
@HeikoS | lambday: nice! | 13:32 |
lambday | parameter - added | 13:32 |
lambday | serialization - done | 13:33 |
lambday | modular typemap - left :( | 13:33 |
@HeikoS | got a meeting now, be back in a bit | 13:33 |
lambday | HeikoS: okay.. see you :) | 13:33 |
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 13:45 | |
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has joined #shogun | 13:57 | |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has joined #shogun | 14:02 | |
vgorbati | HeikoS: hello, a few weeks ago you gave a link to the book about bayesian approach and machine learning to one guy. I have lost the link accidentally, could you please paste it one more time? | 14:04 |
-!- iglesiasg [c1934d16@gateway/web/freenode/ip.193.147.77.22] has quit [Ping timeout: 250 seconds] | 14:05 | |
sonne|work | lambday: sounds very good! | 14:05 |
lambday | sonne|work: I added few useless stuffs too... will remove it | 14:09 |
lambday | sonne|work: will just be back after having some food.. didn't have breakfast, didn't have lunch today :( | 14:09 |
sonne|work | errm well get some! | 14:09 |
lambday | brb | 14:10 |
-!- foulwall [~foulwall@2001:da8:215:503:b82a:3f21:307:6de8] has joined #shogun | 15:07 | |
foulwall | ping sonney2k, sonne|work | 15:08 |
@HeikoS | lambday, vgorbati back | 15:19 |
@HeikoS | vgorbati: did you mean the GP for machine learning book? | 15:19 |
@HeikoS | or the david barber one? | 15:20 |
vgorbati | HeikoS: no, I remember it was bayesian (at least at the beginning of the book) | 15:20 |
@HeikoS | Bayesian Reasoning and Machine Learning? | 15:20 |
vgorbati | HeikoS: could you please give a link, I'll check then) | 15:20 |
@HeikoS | google for this title and append "David Barber" | 15:20 |
vgorbati | HeikoS: ok, thanks | 15:20 |
@HeikoS | is it the one? | 15:20 |
vgorbati | HeikoS: yes, thanks a lot, it is really a good one | 15:21 |
@HeikoS | yep its quite nice | 15:21 |
@HeikoS | though sometimes a bit complicated | 15:22 |
@lisitsyn | HeikoS: complicated for you? I should not get to that book ever | 15:28 |
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has joined #shogun | 15:45 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 15:45 | |
-!- nube [~rho@49.244.69.116] has joined #shogun | 15:45 | |
@HeikoS | lisitsyn: I prefer simple things! | 15:48 |
-!- foulwall [~foulwall@2001:da8:215:503:b82a:3f21:307:6de8] has quit [Remote host closed the connection] | 15:51 | |
lambday | HeikoS: sorry I was away | 15:54 |
@HeikoS | votjakovr: nice with your blog! | 15:54 |
votjakovr | HeikoS: Hi, thanks :) | 15:56 |
@HeikoS | votjakovr: did you manage to do all the tax-form stuff? | 16:00 |
@lisitsyn | what tax-form :D | 16:00 |
@HeikoS | lisitsyn: gsoc money | 16:01 |
votjakovr | HeikoS: yep, i sent it | 16:01 |
@HeikoS | votjakovr: nice, how is the state on the re-structuring? | 16:01 |
@lisitsyn | ahhh that thing | 16:01 |
@lisitsyn | not really a tax thing | 16:01 |
@lisitsyn | HeikoS: btw did you have to pay some taxes from your gsoc $? | 16:02 |
@HeikoS | lisitsyn: nope | 16:02 |
@HeikoS | no income in Germany | 16:02 |
@lisitsyn | HeikoS: good | 16:02 |
@HeikoS | but did it from Germany | 16:02 |
@lisitsyn | ehm? | 16:02 |
@lisitsyn | so did you? | 16:03 |
@HeikoS | lisitsyn: ehm sorry typo | 16:03 |
@HeikoS | no of course not | 16:03 |
@HeikoS | I was here | 16:03 |
@HeikoS | first time in Germany | 16:03 |
@HeikoS | never paid anything | 16:03 |
@lisitsyn | got it | 16:03 |
@lisitsyn | HeikoS: I am just curious how is it done in first world countries :D | 16:04 |
@HeikoS | very complicated everything | 16:04 |
votjakovr | HeikoS: I'm working on it... I added enum for inference type casting, moved latent features to FITC, removed m_data_means and m_label_vector from InferenceMethod class | 16:04 |
lambday | its just one page saying that I didn't do anything while residing in the US, right? | 16:04 |
@HeikoS | votjakovr: that sounds good! | 16:04 |
@HeikoS | lambday: yes | 16:05 |
lambday | hmmm | 16:05 |
lambday | HeikoS: by the way, I am not sure about the modular typemaps | 16:05 |
lambday | do I have to do something with the swig files or something? | 16:05 |
@HeikoS | votjakovr: what about this pulling a general GP class out from the regression framework? You wanted to do this a bit later right? | 16:06 |
@HeikoS | lambday: thats easy | 16:06 |
votjakovr | HeikoS: yep | 16:06 |
@HeikoS | lambday: have a look at the files for python | 16:07 |
@HeikoS | every class has to be added - done | 16:07 |
lambday | HeikoS: okay I'm checking | 16:07 |
@HeikoS | to add a new type, talk to lisitsyn or sonney2k, sonne|work I have never done this, but it should be straight foward | 16:07 |
@lisitsyn | what type | 16:07 |
@lisitsyn | ? | 16:07 |
@HeikoS | votjakovr: I suggest this to be next after the current changes | 16:07 |
lambday | complex64_t | 16:07 |
@HeikoS | lisitsyn: compley numbers | 16:07 |
@lisitsyn | for features? | 16:08 |
lambday | python has complex double or something? | 16:08 |
@HeikoS | lisitsyn: no | 16:08 |
lambday | nah for matrices and vectors | 16:08 |
@HeikoS | lambday: To be honest, I dont know ;) chceking | 16:08 |
lambday | HeikoS: neither do I :( | 16:08 |
@HeikoS | lambday: | 16:08 |
@lisitsyn | python can haz | 16:08 |
@lisitsyn | complex | 16:08 |
@HeikoS | start ipython | 16:08 |
@HeikoS | --pylab | 16:08 |
@HeikoS | and do complex<tab> | 16:08 |
votjakovr | HeikoS: I agree :) | 16:09 |
@HeikoS | youll see numpy has all these types | 16:09 |
@HeikoS | In [3]: complex64(1) | 16:09 |
@HeikoS | Out[3]: (1+0j) | 16:09 |
@lisitsyn | HeikoS: IIRC built-in | 16:09 |
@lisitsyn | just 'j' | 16:09 |
@HeikoS | lisitsyn: yes its numpy | 16:09 |
@lisitsyn | no, not numpy I think | 16:09 |
@lisitsyn | built-in | 16:09 |
@HeikoS | votjakovr: cool, this should be it for framework specific changes then, afterwards, we could start drawing diagrams of the new classification parts | 16:09 |
@HeikoS | lisitsyn: yes, but numpy adds different sizes | 16:10 |
@HeikoS | In [5]: complex64 | 16:10 |
@HeikoS | Out[5]: numpy.complex64 | 16:10 |
@HeikoS | In [6]: complex128 | 16:10 |
@HeikoS | Out[6]: numpy.complex128 | 16:10 |
lambday | HeikoS: NameError: name 'complex64' is not defined | 16:10 |
lambday | :( | 16:10 |
@HeikoS | lambday: did you start with --pylab? | 16:10 |
lambday | yeah | 16:10 |
@HeikoS | from numpy import complex64 | 16:11 |
votjakovr | HeikoS: yep, I think it's better to draw diagram for whole framework including regression part. | 16:11 |
@HeikoS | what does that give? | 16:11 |
lambday | oh yeah | 16:11 |
lambday | that works :) | 16:11 |
lambday | yay | 16:11 |
@HeikoS | votjakovr: yes, good idea, but try to keep it simple, since this is only about getting a big picture overview (no getters/setters, too many variables etc) | 16:11 |
@HeikoS | votjakovr: only things like features, etc should be in there | 16:12 |
lambday | okay I'm checking python_modular swig_typemaps.i then | 16:12 |
@HeikoS | lambday: yes, cool | 16:13 |
@HeikoS | lambday: note that the typemaps are not priority for now | 16:13 |
votjakovr | HeikoS: Ok | 16:13 |
@HeikoS | as long as you did not add your classes to swig, everything will work fine | 16:13 |
@HeikoS | (though not be available for swig for now= | 16:13 |
@HeikoS | we can do these things later | 16:13 |
lambday | HeikoS: okay... that sounds better | 16:13 |
@HeikoS | when all modular stuff is added | 16:14 |
@HeikoS | then you can do it all at once | 16:14 |
lambday | HeikoS: that's better | 16:14 |
@HeikoS | so I would suggest, if the c++ parts work, lets press on | 16:14 |
lambday | HeikoS: okay.. so I'm cleaning up, and submitting a PR | 16:14 |
@HeikoS | lambday: nice! | 16:14 |
lambday | (this time its really gonna look ugly :(, its so big and so many things changed, even I lost track.. thanks to git it remembers everything :D ) | 16:15 |
@HeikoS | lambday: yes, but its hard to do this in little steps | 16:15 |
@HeikoS | make sure that it will be sent as one commit | 16:15 |
lambday | HeikoS: alright | 16:15 |
@HeikoS | so that the buildbot doesn not fail on al lthe sub-commits | 16:15 |
@HeikoS | and make sure unit-tests of new code are included, | 16:15 |
@HeikoS | this makes it easier | 16:15 |
lambday | okay | 16:15 |
lambday | I added few useless stuffs in CMath, wasn't sure about those.. some useless SG_ERROR msgs... I think we don't really need that.. | 16:16 |
@HeikoS | such as? | 16:16 |
lambday | what you said, that it gives error in compile time. should be enough for the user | 16:16 |
lambday | right? | 16:16 |
@HeikoS | I see, so the error messages are in fact never produced since one cannot compile the code? | 16:17 |
lambday | no no | 16:17 |
lambday | it compiles... but when you use that method it shows the error message | 16:17 |
@HeikoS | thats very useful! | 16:17 |
lambday | but it that really needed? since its not a template class | 16:17 |
@HeikoS | lambday: I dont get it | 16:18 |
lambday | HeikoS: CMath has few static template methods... I wrote many error msgs for complex64_t for many of them... | 16:19 |
@HeikoS | ok | 16:19 |
@HeikoS | thats all good | 16:19 |
lambday | okay I'll send the PR... then you can have a look | 16:19 |
lambday | not removing anything then | 16:19 |
lambday | by the way, if I configure shogun with --disable-eigen3, make unit-tests gives compilation errors :-/ | 16:20 |
@HeikoS | lambday: which ones? | 16:21 |
lambday | SGObject | 16:22 |
@HeikoS | ehm? | 16:23 |
@HeikoS | really | 16:23 |
lambday | wait | 16:23 |
lambday | https://gist.github.com/lambday/5698514 | 16:24 |
lambday | really weird | 16:25 |
@HeikoS | well, probably some problems with #defines and stuff, GaussianKernel not declard etc | 16:26 |
@HeikoS | will have a look later | 16:27 |
lambday | HeikoS: okay.. | 16:27 |
lambday | I have added one unit test for math complex stuffs, another for sgvector complex stuffs | 16:28 |
lambday | rest, haven't written yet | 16:28 |
@HeikoS | lambday: ok, we can already merge, but please do the tests next then | 16:28 |
@HeikoS | I know its a little annoying but these are very fundamental things which made sure to work | 16:29 |
lambday | HeikoS: yes.. I was thinking to add those now in fact, instead of doing it later... | 16:29 |
@HeikoS | submit the patch first and then do them | 16:30 |
@HeikoS | otherwise it will get too messy | 16:30 |
lambday | HeikoS: alright.. okay :) | 16:30 |
lambday | checking and cleaning up then... | 16:30 |
@iglesiasg | someone has compiled with directors enabled? I think it is broken | 16:37 |
@iglesiasg | it is an easy fix, just wanted to check if someone found it as well | 16:37 |
@HeikoS | iglesiasg: I dont have compiled this sorry | 16:38 |
@HeikoS | iglesiasg: have you used the directors stuff yet? | 16:39 |
@iglesiasg | HeikoS: yeah, all the time for my thesis :) | 16:39 |
@iglesiasg | I implemented a structured model in python | 16:39 |
@HeikoS | iglesiasg: cool!, maybe you can tell me a bit about it? what do you use it for? | 16:39 |
@HeikoS | I see | 16:39 |
@iglesiasg | HeikoS: I can put you the example where I use it | 16:39 |
@iglesiasg | so I have my StructuredOutputMachine | 16:40 |
@HeikoS | so I want to do something, but I am not too sure whether its possible | 16:40 |
@HeikoS | but yeah, explain first :) | 16:40 |
@iglesiasg | that it basically solves an optimization problem for structured SVMs | 16:40 |
@iglesiasg | this problem has some application dependent parts | 16:40 |
@iglesiasg | there are a couple of functions that change if you are dealing with e.g. sequences, or graphs, or other kind of structured objects | 16:41 |
@iglesiasg | these application dependent functions are encapsulated in a structured model | 16:41 |
@iglesiasg | so using director classes | 16:41 |
@iglesiasg | I can implement a structured model in a target language | 16:41 |
@iglesiasg | and the code in the StructuredOutputMachine is "calling" this code in the target language | 16:42 |
@HeikoS | so one can have an abstract method, that is being called from a c++ implementation | 16:42 |
@HeikoS | but this method is actually implemented in say python | 16:42 |
@iglesiasg | yes | 16:42 |
@HeikoS | and then I can call the c++ implementation from octave | 16:42 |
@HeikoS | which then calls the python code | 16:43 |
@HeikoS | ? | 16:43 |
@iglesiasg | I haven't tried the case of calling from octave, but yes it should work | 16:43 |
@HeikoS | I meant just any other language | 16:43 |
@iglesiasg | yeah | 16:43 |
@HeikoS | this is awesome | 16:43 |
@HeikoS | so what I want to do is | 16:43 |
@iglesiasg | pretty cool yeah | 16:43 |
@iglesiasg | we should eternally thank gsomix :) | 16:43 |
@HeikoS | to write a general MCMC framework ( to generate samples from distributions) | 16:44 |
@HeikoS | for this, there will be a lot of code | 16:44 |
@HeikoS | but obviously, we cannot code up every distribution one wants to sample | 16:44 |
gsomix | iglesiasg: wut? | 16:44 |
gsomix | ah, ok | 16:44 |
@HeikoS | and the c++ code would need to call a method called say log_pdf() | 16:44 |
@HeikoS | but this has to be a user provided unfction | 16:45 |
@HeikoS | function | 16:45 |
@HeikoS | but I dont want to make people code this up in c++ | 16:45 |
@HeikoS | so they can do it in pytho | 16:45 |
@HeikoS | but things would still work | 16:45 |
@HeikoS | how hard is it to do that? | 16:45 |
@HeikoS | so if I have a base class CMCMCTarget | 16:45 |
@HeikoS | which has some abstract methods | 16:45 |
@HeikoS | what do I need to do in order to implement them in python | 16:46 |
@iglesiasg | so I know the C++ part | 16:46 |
@iglesiasg | you need to do a Director class, which is pretty straightforward | 16:46 |
@iglesiasg | and then your python class extends this director class | 16:46 |
@iglesiasg | see for example DirectorDotFeatures | 16:46 |
@HeikoS | does the user has to write it? | 16:47 |
@iglesiasg | no | 16:47 |
@HeikoS | nice | 16:47 |
@iglesiasg | in your case you will need to do a | 16:47 |
@iglesiasg | DirectorMCMCTarget | 16:47 |
@iglesiasg | HeikoS: the thing is that this probably requires some SWIG code, I am not sure though | 16:48 |
@iglesiasg | gsomix: for directors, you need the director class in C++ and SWIG code as well or just the director? | 16:48 |
@HeikoS | ok, but I can prepare everything and the user just has to extend an existing python class with the method? | 16:49 |
@iglesiasg | HeikoS: yes, that is the workflow | 16:49 |
gsomix | iglesiasg: just director class in C++. | 16:49 |
@iglesiasg | gsomix: good | 16:49 |
@HeikoS | gsomix, iglesiasg, awesome! :) | 16:49 |
@HeikoS | this is exactly what I needed | 16:50 |
@iglesiasg | HeikoS: just a remark | 16:50 |
@HeikoS | from which languages does this work? | 16:50 |
@HeikoS | and how expensive is it? | 16:50 |
@iglesiasg | HeikoS: you don't actually extend an existing python class, you extend an existing C++ class from python | 16:50 |
@HeikoS | I see, | 16:50 |
@iglesiasg | HeikoS: that's interesting, I wanted to see how slow was it from say pure C++ but never tried it at the end | 16:51 |
@iglesiasg | if you meant that with expensive | 16:51 |
@HeikoS | iglesiasg: this probably depends on the method | 16:52 |
@HeikoS | in general I mean | 16:52 |
@HeikoS | but I was interested in the amount of overhead | 16:52 |
@iglesiasg | definitely | 16:52 |
@iglesiasg | no idea | 16:52 |
@iglesiasg | lisitsyn mentioned once that it could be slow | 16:52 |
gsomix | very slow | 16:54 |
gsomix | cu later | 16:54 |
-!- gsomix [~Miranda@r206-10.smr.ru] has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org] | 16:54 | |
@HeikoS | iglesiasg: ok, very slow :) | 16:54 |
@iglesiasg | hehe | 16:54 |
votjakovr | HeikoS: i have a question: i'd like to scale SGMatrix (simple multiply by a real number), what is the best way for doing this? | 16:57 |
@HeikoS | votjakovr: either there is already a method to do this. If not, you could create one, use eigen's Map, or just write the loop | 16:58 |
@HeikoS | having a method scale(T scalar) would be probably good anyway | 16:59 |
sonne|work | votjakovr: just make sure you have the implementation only in the .cpp file and no dependency to eigen in the .h | 17:00 |
@HeikoS | votjakovr, sonne|work if you add this to SGMatrix, then better not use eigen anyway | 17:00 |
votjakovr | HeikoS: Ok, I'll add it | 17:00 |
@HeikoS | sonne|work: btw I really dont like our matrix representation | 17:00 |
sonne|work | gtg | 17:01 |
@lisitsyn | HeikoS: what do you don't like? | 17:02 |
@HeikoS | lisitsyn: its not oop | 17:02 |
@HeikoS | both matrix and vector should be scalable | 17:02 |
@HeikoS | for example | 17:02 |
@lisitsyn | scalable? | 17:02 |
@HeikoS | so only one method | 17:02 |
@HeikoS | multiply by scalar | 17:02 |
@HeikoS | but this is not reflecte | 17:02 |
@HeikoS | d | 17:02 |
@lisitsyn | ahh | 17:02 |
@lisitsyn | I would not do anything like that in sgmatrix | 17:03 |
@lisitsyn | I don't like we put such methods here | 17:03 |
@lisitsyn | as I consider it as just a wrapper for swig deals and storages | 17:03 |
@HeikoS | lisitsyn: agreed! | 17:03 |
@HeikoS | votjakovr: then please dont add this method | 17:03 |
@lisitsyn | I'd rather add a method | 17:03 |
@lisitsyn | that returns eigen map | 17:03 |
@lisitsyn | for that matrix | 17:03 |
@HeikoS | votjakovr: ^ | 17:04 |
@HeikoS | good idea | 17:04 |
@HeikoS | we can in fact then also start to remove all these matrix functions from there (mean/covariance etc) | 17:04 |
@lisitsyn | and then we do matrix.map().whatever we ever wanted() | 17:04 |
@HeikoS | lisitsyn: thats much better | 17:04 |
@HeikoS | then I will like the current classes again :D | 17:04 |
@lisitsyn | hah alright | 17:04 |
@HeikoS | you are right, they are rather wrappers | 17:05 |
@HeikoS | gotta go now, see you later | 17:05 |
@lisitsyn | me too | 17:05 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 17:05 | |
votjakovr | HeikoS: so you think, that it is better to use eigen3 for that purposes? | 17:07 |
@HeikoS | votjakovr: yes I strongly agree with sergey | 17:07 |
@HeikoS | the SGMatrix class should have a method to return an eigen3 matrix | 17:07 |
@HeikoS | and then you can use the usual operations on that | 17:07 |
@HeikoS | then code is very clean and we are not re-inventing the wheel all the time | 17:08 |
@HeikoS | votjakovr: https://github.com/shogun-toolbox/shogun/issues/1139 | 17:12 |
votjakovr | HeikoS: oh yep, i agree, it will be great | 17:12 |
@HeikoS | votjakovr: nice! :) | 17:12 |
@HeikoS | gotta go now,see you! | 17:12 |
-!- HeikoS [~heiko@nat-160-98.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 17:12 | |
votjakovr | HeikoS: cu | 17:12 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 17:21 | |
shogun-notifier- | shogun: iglesias :develop * 8496e3f / src/shogun/ (13 files): https://github.com/shogun-toolbox/shogun/commit/8496e3f75c3a73073815f6ede5c64360f93de4a3 | 17:21 |
shogun-notifier- | shogun: Add support for continuous features or observations to the HMSVM | 17:21 |
shogun-notifier- | shogun: - Major changes in HMSVMModel | 17:21 |
shogun-notifier- | shogun: - Add convenience init_training method in StructuredModel | 17:21 |
shogun-notifier- | shogun: - TwoStateModel::simulate_two_state_data -> simulate_data with arguments | 17:21 |
shogun-notifier- | shogun: - Fix bug with default value of m_use_director_risk in StructuredModel cto. | 17:21 |
shogun-notifier- | shogun: iglesias :develop * f4f4e2c / examples/undocumented/python_modular/ (6 files): https://github.com/shogun-toolbox/shogun/commit/f4f4e2c2f53b42460348efcf29d784c9cfc1b5a9 | 17:21 |
shogun-notifier- | shogun: Rename old HMSVM python examples to discrete and add new ones for PLiFs | 17:21 |
shogun-notifier- | shogun: iglesias :develop * 64e789a / / (22 files): https://github.com/shogun-toolbox/shogun/commit/64e789a970759437e078e5cbdeedb685148e4ef2 | 17:21 |
shogun-notifier- | shogun: Several minor improvements in the SO framework and the HMSVM | 17:21 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 923f9b3 / / (33 files): https://github.com/shogun-toolbox/shogun/commit/923f9b343f4ea81b7663e14eaa40bf0bba17f75e | 17:21 |
shogun-notifier- | shogun: Merge pull request #1089 from iglesias/feature/hmsvm_plifs | 17:21 |
shogun-notifier- | shogun: | 17:21 |
shogun-notifier- | shogun: Support for continuous features with the HM-SVM using PLiFs | 17:21 |
-!- vgorbati [c3ee5cb1@gateway/web/freenode/ip.195.238.92.177] has quit [Ping timeout: 250 seconds] | 17:29 | |
-!- gsomix [~gsomix@185.6.239.124] has joined #shogun | 17:30 | |
gsomix | good evening | 17:34 |
-!- yoo [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has joined #shogun | 17:37 | |
yoo | hiall | 17:37 |
-!- yoo [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has quit [Quit: Page closed] | 17:48 | |
-!- foulwall [~foulwall@2001:da8:215:c252:8038:73e7:c08a:637a] has joined #shogun | 17:50 | |
-!- lisitsyn [~blackburn@185.6.239.124] has joined #shogun | 17:58 | |
lisitsyn | http://izbicki.me/blog/hlearn-cross-validates-400x-faster-than-weka | 17:59 |
@iglesiasg | hehe | 18:03 |
@iglesiasg | that seems pretty exaggerated | 18:03 |
@iglesiasg | aah it is another algorithm | 18:04 |
@iglesiasg | and they are using there >100 folds | 18:04 |
@iglesiasg | lisitsyn: is it normal to use so many? I have normally gone with 5 hehe | 18:05 |
lisitsyn | iglesiasg: noooo idea! | 18:05 |
lisitsyn | well 1000 folds is ehm | 18:05 |
lisitsyn | pretty much yes | 18:06 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 18:15 | |
-!- iglesiasg [c1934d18@gateway/web/freenode/ip.193.147.77.24] has quit [Quit: Page closed] | 18:22 | |
-!- iglesiasg [~iglesiasg@193.147.77.24] has joined #shogun | 18:25 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 18:25 | |
-!- foulwall [~foulwall@2001:da8:215:c252:8038:73e7:c08a:637a] has quit [Remote host closed the connection] | 18:43 | |
van51 | sonney2k: are you here | 19:34 |
van51 | ? | 19:34 |
-!- hushell [~hushell@c-67-189-100-116.hsd1.or.comcast.net] has joined #shogun | 19:37 | |
-!- iglesiasg [~iglesiasg@193.147.77.24] has quit [Quit: Leaving] | 19:46 | |
@sonney2k | van51, hey! | 19:47 |
van51 | sonney2k: hello! | 19:47 |
van51 | sonney2k: first of all, I edited the PR | 19:48 |
van51 | in regards to your comments | 19:48 |
@sonney2k | lambday, you are a superhero! | 19:48 |
van51 | secondly, we managed to meet today :D | 19:48 |
van51 | and we said we continue our interactions mainly through emails at least with Olivier and Benoit | 19:49 |
van51 | and they wanted to know whether we would have to continue CC'ing you | 19:49 |
van51 | not that they mind or anything | 19:49 |
lambday | sonney2k: watching the travis build like I watch a status bar while downloading.. :( | 19:49 |
van51 | lambday: I do that too! | 19:50 |
lambday | van51: :) | 19:50 |
@sonney2k | lambday, I had a look over your code. It is good to merge from my side! | 19:50 |
@sonney2k | lambday, so as soon as the bb turns green I will merge | 19:50 |
lambday | sonney2k: thanks | 19:51 |
lambday | sonney2k: if this works, I'm gonna give myself a huge treat... some fancy restaurants for sure :-/ | 19:51 |
lambday | sonney2k: the macro idea was a life-saver | 19:53 |
@sonney2k | van51, regarding the PR we still need a unit test (again) and yes you are right I have to find some example | 19:53 |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has quit [Read error: No route to host] | 19:56 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 19:58 | |
@sonney2k | van51, regarding the PR we still need a unit test (again) and yes you are right I have to find some example | 20:00 |
lambday | sonney2k: something failed :( | 20:05 |
lambday | sonney2k: oh its just cause of high precision | 20:05 |
lambday | sonney2k: should I change it now? in the unit-test I mean? | 20:07 |
@sonney2k | lambday, yes precision of double is just 1e-16 | 20:07 |
@sonney2k | so don't go above that | 20:07 |
lambday | sonney2k: okay | 20:07 |
@sonney2k | 1e-15 is reasonable | 20:07 |
lambday | sonney2k: okay.. I'll change it | 20:07 |
lisitsyn | sonney2k: do you agree with our thoughts on sgmatrix? | 20:08 |
lambday | ah, finally! :D | 20:18 |
lambday | sonney2k: should I change it in this PR or in the next PR? I'll be editing CJacobiEllipticFunctions in the next PR with this complex64_t.. | 20:19 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 20:21 | |
@sonney2k | lambday, just change it in this PR and force push. I am currently checking why it is deviating at all though | 20:22 |
lambday | sonney2k: alright | 20:22 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 20:25 | |
shogun-notifier- | shogun-demo: foulwall :master * b16be7b / / (10 files): https://github.com/shogun-toolbox/shogun-demo/commit/b16be7bfa4389e8e14214215dfdaf5b34c19e740 | 20:25 |
shogun-notifier- | shogun-demo: Simplify the argument panel. when creating new demos, developer only | 20:25 |
shogun-notifier- | shogun-demo: need to fillin a python dict with types and names of each argument. | 20:25 |
shogun-notifier- | shogun-demo: Soeren Sonnenburg :master * 05615f2 / / (10 files): https://github.com/shogun-toolbox/shogun-demo/commit/05615f2c7b988ca9b137f9ffc2c123a4399903b7 | 20:25 |
shogun-notifier- | shogun-demo: Merge pull request #5 from foulwall/refractor | 20:25 |
shogun-notifier- | shogun-demo: | 20:25 |
shogun-notifier- | shogun-demo: Simplify the argument panel. | 20:25 |
lambday | sonney2k: changed | 20:31 |
lambday | argh! | 20:39 |
@sonney2k | lambday, ok it seems this NEAR stuff is for double only | 20:44 |
@sonney2k | so don't try to be overly accurate | 20:44 |
lambday | sonney2k: okay.. | 20:45 |
lambday | but in many cases I've kept it as 1e-19, whenever it passed | 20:45 |
shogun-notifier- | shogun: lambday :develop * 97f4e8f / / (22 files): https://github.com/shogun-toolbox/shogun/commit/97f4e8f6a1ed7d1037a29b974885116619f35875 | 20:45 |
shogun-notifier- | shogun: ptype complex64_t added | 20:45 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * bc4689c / / (22 files): https://github.com/shogun-toolbox/shogun/commit/bc4689c0c06768661d35cf137f883b14348aa5ff | 20:45 |
shogun-notifier- | shogun: Merge pull request #1140 from lambday/develop | 20:45 |
shogun-notifier- | shogun: | 20:45 |
shogun-notifier- | shogun: ptype complex64_t added | 20:45 |
@sonney2k | lambday, alright then | 20:45 |
lambday | sonney2k: thank you :) :) | 20:46 |
lambday | phew! | 20:47 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 2e6b11f / examples/undocumented/ (3 files): https://github.com/shogun-toolbox/shogun/commit/2e6b11fc1804435c3db7dd1c5fc078301369d31e | 20:47 |
shogun-notifier- | shogun: drop useless unused includes | 20:47 |
@sonney2k | van51, there is actually no committed example using that :/ | 20:47 |
@sonney2k | van51, I only had some research code written like about 4 years ago so all we can rely on is your tests for now | 20:48 |
@sonney2k | van51, however whenever you make some progress with hashing we can resurrect the old code and make it a real example | 20:49 |
@sonney2k | van51, ahh and yes please keep me in the CC if possible thanks! | 20:49 |
van51 | sonney2k: ok sounds like a plan :) | 20:49 |
van51 | we will! | 20:49 |
van51 | although with quoc it will probably be gtalk conversations | 20:50 |
@sonney2k | van51, ok then please ping me when you have updated the PR | 20:50 |
@sonney2k | van51, yeah do as you can but keep me updated then | 20:50 |
van51 | yeah ofc | 20:52 |
shogun-buildbot | build #894 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/894 blamelist: lambday <heavensdevil6909@gmail.com> | 20:58 |
lambday | failed?? | 20:58 |
shogun-buildbot | build #1062 of bsd1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1062 blamelist: lambday <heavensdevil6909@gmail.com> | 20:58 |
@sonney2k | lambday, ahh cyg ... | 21:00 |
@sonney2k | hmmhh bsd?! | 21:00 |
* sonney2k checks | 21:00 | |
lambday | yeah bsd :( | 21:00 |
@sonney2k | wiking, http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1062/steps/compile/logs/stdio | 21:01 |
@sonney2k | lambday, is complex standard stuff? | 21:01 |
lambday | sonney2k: yes! | 21:01 |
lisitsyn | why real(c_real)? | 21:02 |
shogun-buildbot | build #1063 of bsd1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1063 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:02 |
lambday | lisitsyn: to set the real and imag | 21:03 |
lambday | it worked here | 21:03 |
lisitsyn | lambday: why not real() = c_real? | 21:04 |
lisitsyn | lambday: i have such methods in my std lib too | 21:04 |
lisitsyn | not sure about bsd | 21:04 |
lambday | lisitsyn: bad :( | 21:05 |
@sonney2k | lambday, real(val) and imag(val) are C++11 extensions | 21:05 |
lambday | I'm gonna have to change those then... :( | 21:05 |
@sonney2k | lambday, as are proj(), asin(), acos(), a* | 21:05 |
lambday | sonney2k: so we shouldn't be using them? | 21:06 |
@sonney2k | lambday, well we could have a configure check testing for it | 21:06 |
@sonney2k | but I am afraid it won't save us | 21:06 |
lambday | sonney2k: asin, acos thing we need for the project :( | 21:06 |
lisitsyn | sonney2k: can't see where is it stated | 21:06 |
@sonney2k | lisitsyn, http://books.google.de/books?id=9DEJKhasp7gC&pg=PT1310&lpg=PT1310&dq=c%2B%2B+complex+standard&source=bl&ots=_917L8IrRD&sig=EqmCPSeGuAsrFivfrVgA4lqLICU&hl=en&sa=X&ei=pOisUZHxKouqPJfEgKAE&redir_esc=y | 21:07 |
lisitsyn | sonney2k: well real() returns reference | 21:08 |
lisitsyn | we still can real() = 3 I think? | 21:08 |
@sonney2k | well we have 3 options: 1) we require c++11 compilers | 21:09 |
@sonney2k | 2) we don't use st' | 21:10 |
@sonney2k | stl's complex type but one from some other lib | 21:10 |
lisitsyn | sonney2k: what is wrong with older std? | 21:10 |
@sonney2k | 3) we work around it | 21:10 |
@sonney2k | asin, acos etc missing which lambday needs | 21:10 |
lisitsyn | ah okay | 21:11 |
lambday | make it c++11 compilers :( | 21:11 |
lambday | configure won't help? | 21:11 |
@sonney2k | hmmhh, all difficult. shogun currently works with zero dependencies | 21:12 |
gsomix | sonney2k, tnx for review. working... | 21:12 |
@sonney2k | just the parts that are unavailable (like blas/lapack/eigen/nlopt/ etc etc) won't work | 21:13 |
shogun-buildbot | build #895 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/895 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:13 |
@sonney2k | I mean parts depending on the libs that are unavailable won't work | 21:13 |
@sonney2k | but that is ok | 21:13 |
lambday | sonney2k: I'm sorry... we don't need asin, acos.. | 21:13 |
lambday | sorry for the confusion | 21:13 |
@sonney2k | what researchers I met most complained about shogun back then was | 21:14 |
@sonney2k | that they couldn't get it compiled(!) | 21:14 |
@sonney2k | too complex. too difficult to get to work | 21:14 |
@sonney2k | so I am not too fond of creating hard dependencies | 21:14 |
lambday | sonney2k: okay... but I just checked and we don't need asin, acos, atan etc... | 21:14 |
@sonney2k | lambday, ok perfect :D | 21:15 |
lambday | so, if the serialization thing can be taken care of | 21:15 |
@sonney2k | ? | 21:15 |
lambday | I'll check the real() = some_val thing that lisitsyn was sayng | 21:15 |
lambday | if that works, then we're good I guess? | 21:15 |
lambday | sonney2k: its easy to change I guess | 21:20 |
@sonney2k | lambday, yes it works | 21:21 |
@sonney2k | lambday, so just change it to x.real() = foo | 21:21 |
lambday | sonney2k: okay | 21:21 |
lambday | changing.. | 21:21 |
@sonney2k | gsomix, I still don't understand your code... | 21:22 |
@sonney2k | gsomix, could you please explain how circular buffer works | 21:22 |
@sonney2k | I mean what is write_to_buffer doing? | 21:23 |
lambday | sonney2k: changed.. sending PR | 21:28 |
gsomix | sonney2k, method just locates free memory in buffer and copies data to it. logically just append array of bytes to end of buffer. but free memory can be situated as at end and at begin of physical memory, because buffer is ring (just imagine the tape with glued ends). | 21:32 |
gsomix | sonney2k, I hope this helped | 21:32 |
lambday | my hostel has this weird policy since last 4-5 days that exactly 1:30 at night (20 UTC) my internet gets disconnected.. comes back 3:30 hrs later | 21:34 |
@sonney2k | gsomix, hmmhh so you do harddisk reads in smaller blocks then right? | 21:36 |
gsomix | sonney2k, yep. size of blocks is configurable | 21:37 |
@sonney2k | I still don't have the full picture | 21:37 |
@sonney2k | so lets start an example | 21:38 |
@sonney2k | I read in say 1MB of data from the file | 21:38 |
@sonney2k | into my circular buffer (I hope) | 21:38 |
@sonney2k | then go through the buffer until I find no more '\n' | 21:39 |
@sonney2k | say 10KB before the end of the buffer | 21:39 |
@sonney2k | gsomix, then I do what? | 21:39 |
@sonney2k | and my circular buffer is fixed size 1 MB here right?! | 21:40 |
gsomix | sonney2k, just read another 1MB of data. then 10KB copies into end of physical memory of buffer, and 990KB at begin. | 21:40 |
gsomix | sonney2k, yep | 21:40 |
gsomix | errr | 21:40 |
gsomix | 990 copies at begin | 21:40 |
gsomix | so... you should reat 990K from file | 21:41 |
gsomix | *read | 21:41 |
gsomix | and write to buffer | 21:41 |
gsomix | because available only 990KB of free space in buffer. logical begin of buffer then at (end of physical block - 10KB) position. | 21:43 |
@sonney2k | gsomix, so you read as much as you need to fill in the 1MB | 21:43 |
gsomix | sonney2k, no more than chunk_size (maximal size of block to read) | 21:45 |
@sonney2k | gsomix, ok I see. lets say now you have still a 1MB buffer but the string is 500k long and you don't find another string in there | 21:48 |
@sonney2k | so you read again 500k right? | 21:48 |
gsomix | yep | 21:48 |
@sonney2k | and if you don't find anything in there we say ERROR | 21:48 |
@sonney2k | ? | 21:49 |
gsomix | yes. method get_next_line just returns -1. | 21:49 |
@sonney2k | gsomix, btw for the last line one does not need to find a delimiter at the end | 21:51 |
lambday | sonney2k: please have a look at the PR when you're free.. just used it in two places.. rest should be fine I think.. I'll probably get disconnected in 8 mins :( | 21:52 |
@sonney2k | gsomix, I don't get the difference between buffersize and chunksize... | 21:55 |
gsomix | sonney2k, buffer_size is max size of line, chunk_size if max size of block to read. | 21:56 |
shogun-notifier- | shogun: lambday :develop * badadc2 / src/shogun/io/ (2 files): https://github.com/shogun-toolbox/shogun/commit/badadc2a210f87935c703d1d1e8796b77fa84fb4 | 21:56 |
shogun-notifier- | shogun: real/imag(val) replaced by real()/imag()=val | 21:56 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * b290b09 / src/shogun/io/ (2 files): https://github.com/shogun-toolbox/shogun/commit/b290b0905a67657f4e422ecdd1623a1b1ce75feb | 21:56 |
shogun-notifier- | shogun: Merge pull request #1141 from lambday/develop | 21:56 |
shogun-notifier- | shogun: | 21:56 |
shogun-notifier- | shogun: real/imag(val) replaced by real()/imag()=val to make it work on non C++11 compliant compilers | 21:56 |
@sonney2k | gsomix, then please write this in the doxygen docu | 21:57 |
gsomix | sonney2k, ok | 21:57 |
@sonney2k | gsomix, or even better name the parameter like that | 21:57 |
@sonney2k | gsomix, max_line_length | 21:57 |
* gsomix afk | 21:58 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving] | 21:59 | |
-!- lambday [67157f4c@gateway/web/cgi-irc/kiwiirc.com/ip.103.21.127.76] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 22:01 | |
@sonney2k | so there we have lambday's reliably predicted disconnect | 22:01 |
shogun-buildbot | build #1064 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1064 blamelist: lambday <heavensdevil6909@gmail.com> | 22:18 |
shogun-buildbot | build #1188 of deb3 - modular_interfaces is complete: Failure [failed test ruby_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1188 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:19 |
shogun-buildbot | build #896 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/896 blamelist: lambday <heavensdevil6909@gmail.com> | 22:23 |
shogun-buildbot | build #897 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/897 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:23 |
shogun-buildbot | build #1065 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/1065 blamelist: Soeren Sonnenburg <sonne@debian.org> | 22:30 |
-!- iglesiasg [d58f3279@gateway/web/freenode/ip.213.143.50.121] has joined #shogun | 22:34 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 22:34 | |
-!- lisitsyn [~blackburn@185.6.239.124] has left #shogun [] | 22:40 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has left #shogun ["JOIN #shogun"] | 22:42 | |
-!- van51 [~van51@athedsl-408350.home.otenet.gr] has joined #shogun | 22:42 | |
shogun-buildbot | build #1189 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1189 | 23:11 |
gsomix | may the forc^Wvalgrind be with you | 23:20 |
gsomix | sonney2k, so, I have fixed memory leaks. example for LineReader is ready. example for buffer and many_delimiters_read will be tomorrow. | 23:27 |
gsomix | sonney2k, now I need to prepare for pass-fail exam. dsp, dsp... | 23:29 |
@iglesiasg | good luck with it gsomix | 23:57 |
gsomix | iglesiasg, thanks. this is small exam. most interesting will be at the end of June, huh. | 23:59 |
--- Log closed Tue Jun 04 00:00:03 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!