| --- Log opened Mon Jun 06 00:00:00 2011 | ||
| CIA-18 | shogun: Soeren Sonnenburg master * r8b50634 / (6 files in 3 dirs): | 01:42 |
|---|---|---|
| CIA-18 | shogun: put shogun into org.shogun java package, fix configure / make distclean | 01:42 |
| CIA-18 | shogun: targets and adjust java examples - http://bit.ly/j0VQ3H | 01:42 |
| CIA-18 | shogun: Soeren Sonnenburg master * r37508ab / (src/Makefile.template src/configure): add some classpath magic - http://bit.ly/kOd6ne | 02:14 |
| -!- blackburn [~blackburn@188.122.237.128] has joined #shogun | 09:57 | |
| blackburn | ehh.. some issue with mds | 09:58 |
| blackburn | sonney2k: around? | 10:20 |
| @sonney2k | blackburn, yes | 10:24 |
| blackburn | sonney2k: could you please add two line patch for ClassicMDS.cpp? I forgot to #ifdef HAVE_LAPACK it :( | 10:24 |
| @sonney2k | CIA-18, why didn't you say anything? | 10:24 |
| @sonney2k | already done minutes ago | 10:25 |
| blackburn | ah | 10:25 |
| blackburn | hehe | 10:25 |
| blackburn | you have seen it in a mailing list? | 10:25 |
| blackburn | thank you | 10:25 |
| @sonney2k | no idea why sploving didn't fix this himself | 10:25 |
| @sonney2k | yes | 10:25 |
| blackburn | sonney2k: mds could be applied only to distance with no features, is it ok to add it? | 10:27 |
| blackburn | apply_to_distance or so | 10:27 |
| @sonney2k | as you want | 10:27 |
| blackburn | ok | 10:27 |
| blackburn | it makes possible to get features from distance | 10:28 |
| @sonney2k | http://www.h-online.com/news/item/Report-GitHub-more-popular-than-SourceForge-and-Google-Code-1254752.html | 10:34 |
| blackburn | no surprise :) | 10:35 |
| blackburn | github rocks | 10:35 |
| @sonney2k | I am surprised that ruby is the most popular language on github | 10:36 |
| @sonney2k | I am not so surprised that C++ is still most popular overall | 10:36 |
| @sonney2k | https://days2011.scala-lang.org/sites/days2011/files/ws3-1-Hundt.pdf | 10:38 |
| blackburn | ah nevermind there are many ruby frameworks, ruby utilties and so on | 10:38 |
| @sonney2k | says that C++ is still quite a bit faster ... | 10:38 |
| blackburn | but I haven't seen any really serious thing | 10:38 |
| @sonney2k | but hey so is assembly language if you know what you do | 10:38 |
| blackburn | java is f-g slow! heheh | 10:39 |
| blackburn | will be here 13.00 utc | 11:07 |
| -!- blackburn [~blackburn@188.122.237.128] has quit [Quit: Leaving.] | 11:07 | |
| -!- f-x [~shashwat@117.192.200.132] has joined #shogun | 13:12 | |
| -!- blackburn [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has joined #shogun | 13:21 | |
| -!- f-x [~shashwat@117.192.200.132] has quit [Quit: leaving] | 13:38 | |
| -!- f-x [~shashwat@117.192.200.132] has joined #shogun | 13:40 | |
| -!- f-x [~shashwat@117.192.200.132] has quit [Ping timeout: 252 seconds] | 13:47 | |
| -!- f-x [~shashwat@117.192.195.247] has joined #shogun | 14:04 | |
| -!- cwidmer [~quassel@connect.tuebingen.mpg.de] has joined #shogun | 14:11 | |
| blackburn | cwidmer: how are you? ;) | 14:14 |
| -!- f-x [~shashwat@117.192.195.247] has quit [Ping timeout: 240 seconds] | 14:18 | |
| -!- blackburn_ [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has joined #shogun | 14:22 | |
| -!- blackburn [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has quit [Ping timeout: 252 seconds] | 14:24 | |
| -!- blackburn_ is now known as shogun | 14:25 | |
| shogun | heheeh | 14:25 |
| -!- shogun is now known as blackburn | 14:25 | |
| blackburn | damn | 14:25 |
| @sonney2k | blackburn, evil you | 14:25 |
| blackburn | I need to sleep some hehehe | 14:26 |
| cwidmer | hi all | 14:33 |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 14:37 | |
| -!- f-x [~shashwat@117.192.199.159] has joined #shogun | 14:42 | |
| -!- f-x_ [~shashwat@117.192.199.159] has joined #shogun | 14:45 | |
| -!- f-x_ [~shashwat@117.192.199.159] has quit [Client Quit] | 14:46 | |
| -!- mikiobraun [~mikiobrau@squid.ml.tu-berlin.de] has joined #shogun | 14:51 | |
| @sonney2k | http://timeanddate.com/worldclock/city.html?n=37 | 14:55 |
| @sonney2k | 13:30 UTC == 15:30 Germany | 14:55 |
| @sonney2k | It is 13:30 UTC isn't it? | 15:31 |
| @sonney2k | So blackburn, cwidmer, f-x, mikiobraun, VojtechFranc all here and alive? | 15:31 |
| cwidmer | seems so ^^ | 15:31 |
| cwidmer | yep | 15:31 |
| f-x | alive | 15:31 |
| mikiobraun | ye | 15:32 |
| @sonney2k | OK then lets start | 15:32 |
| @sonney2k | Welcome everyone to this 2nd meeting | 15:32 |
| VojtechFranc | hi all | 15:32 |
| @sonney2k | GSoC is now running just a few weeks and I'd like to know how it is going? | 15:32 |
| blackburn | here | 15:33 |
| @sonney2k | Is every student working full time now (I know blackburn still has exams going on so apart from him)? | 15:33 |
| @sonney2k | VojtechFranc, btw do you know where Alesis is? | 15:34 |
| f-x | yes, my college just ended and now i'm working full time | 15:34 |
| @sonney2k | mikiobraun, any idea about sploving? | 15:34 |
| @sonney2k | aka Baozeng. | 15:34 |
| VojtechFranc | Alesis Novik finish his exams the last week and started to work full time. | 15:34 |
| mikiobraun | no, no idea | 15:34 |
| blackburn | i'm working don'tknowhowmuch time :D | 15:34 |
| @sonney2k | No offense - just try to get the job done :) | 15:35 |
| @sonney2k | I just wanted to say a few things when you work with git: | 15:35 |
| @sonney2k | please use git pull --rebase when you merge from master | 15:36 |
| @sonney2k | otherwise if you use git merge - git will assume your tree is the current master and add all the patches of the true master on top of it | 15:36 |
| -!- heiko1 [~heiko@134.91.10.201] has joined #shogun | 15:36 | |
| heiko1 | hi | 15:36 |
| @sonney2k | we had this situation a few times so please just keep this in mind | 15:37 |
| heiko1 | sorry for the delay, traffic | 15:37 |
| @sonney2k | hi heiko | 15:37 |
| @bettyboo | back?! | 15:37 |
| VojtechFranc | sonney2k, would you recommend us some short tutorial for git? I started to read the main documentation but it takes too much time.. | 15:37 |
| heiko1 | the github doc is nice | 15:37 |
| heiko1 | at least for this stuff we're doing here | 15:38 |
| @sonney2k | we did just start - question was whether everyone is now working full time and that everyone should use git pull --rebase instead of git pull | 15:38 |
| @sonney2k | VojtechFranc, there were a few posts on the shogun mailing list | 15:38 |
| cwidmer | here a quite nice overview to get started with git: http://gitimmersion.com/ | 15:38 |
| @sonney2k | these were pretty helpful to me | 15:38 |
| @sonney2k | besides git I'd like to remind that every student is not doing totally isolated work - if minor glitches e.g. compile failures happen please try to fix them (or use git blame to figure out who did it and actively resolve the issue with him). | 15:39 |
| @sonney2k | or work on sth else shogun related and ask | 15:39 |
| @sonney2k | just don't idle | 15:39 |
| @sonney2k | Apart from that I wanted to ask how we could improve in student<->mentor<->admin communication? | 15:40 |
| @sonney2k | or is everyone fine with the current setup? | 15:40 |
| heiko1 | I am fine. | 15:41 |
| f-x | okay with me at the moment. | 15:41 |
| heiko1 | Well we could have a phone talk as planned | 15:41 |
| mikiobraun | works for me | 15:41 |
| cwidmer | works for me as well | 15:41 |
| @sonney2k | And before I forget please don't forget your weekly summary email - it can be really really short (2-3 sentences) but just let us know that you are still working. | 15:42 |
| cwidmer | re git: here is a blog post that talks about the situation that sören metioned - http://gitready.com/advanced/2009/02/11/pull-with-rebase.html | 15:42 |
| @sonney2k | f-x, and everyone else - if you feel that something is not going well say so as early as possible. This project should be fun and not cause frustration. | 15:43 |
| @sonney2k | heiko1, we will | 15:43 |
| f-x | yes, yes.. sure | 15:43 |
| @sonney2k | alright then mid-term eval is coming up in about 1 month | 15:43 |
| heiko1 | At least for me its ok at the moment, wokring full-fime, having fun, knowing what to do | 15:44 |
| @sonney2k | so I think mentors need to define the requirements for successful midterm progress | 15:44 |
| heiko1 | yes, this would be nice | 15:44 |
| -!- blackburn [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has quit [Ping timeout: 252 seconds] | 15:45 | |
| @sonney2k | e.g. sploving (who is doing the language interfaces could have finished java completely including translating examples) | 15:45 |
| mikiobraun | ok | 15:46 |
| @sonney2k | so just think about it - discuss with your student/potential co-mentor and then we will use that | 15:46 |
| @sonney2k | This shouldn't be too hard so lets just have this ready end of this week and please CC me in that email such that I know you have done so. | 15:47 |
| @sonney2k | Any questions concerns? | 15:47 |
| heiko1 | no | 15:47 |
| @sonney2k | Anyone else? | 15:47 |
| f-x | sonney2k: will that mail for me be from you or John? | 15:48 |
| cwidmer | sounds good | 15:48 |
| VojtechFranc | me not | 15:48 |
| f-x | i mean, he has to be informed about it | 15:48 |
| @sonney2k | f-x, since John is the only mentor that is not here I will contact him | 15:48 |
| f-x | ok, thanks | 15:48 |
| f-x | no further questions from me | 15:49 |
| @sonney2k | So then I think we should just schedule the next meeting | 15:49 |
| -!- blackburn [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has joined #shogun | 15:49 | |
| @sonney2k | How about July 4th 13:00 UTC? | 15:49 |
| heiko1 | ok for me | 15:49 |
| VojtechFranc | ok for me | 15:49 |
| blackburn | err connection problem | 15:49 |
| f-x | i can't think of any problems with that date | 15:50 |
| cwidmer | I'll be travelling until July 6th, maybe a few days later? | 15:50 |
| f-x | ^ for me | 15:50 |
| cwidmer | 7th? | 15:50 |
| @sonney2k | would also be fine with me... | 15:50 |
| blackburn | what are the date you talking about? | 15:51 |
| mikiobraun | ok with me | 15:51 |
| heiko1 | also for me | 15:51 |
| f-x | me too | 15:51 |
| @sonney2k | blackburn, next meeting July 7th 13:00 UTC | 15:51 |
| @sonney2k | ok then | 15:51 |
| blackburn | ah I see, it should be ok for me | 15:51 |
| @sonney2k | then thanks everyone for attending the meeting and happy coding / gsoccing | 15:51 |
| heiko1 | thanks, see you! | 15:52 |
| cwidmer | thanks sören | 15:52 |
| @sonney2k | meeting over ... | 15:52 |
| f-x | see you guys | 15:52 |
| cwidmer | happy hacking to you all | 15:52 |
| cwidmer | see you | 15:52 |
| VojtechFranc | thanks, and cheers | 15:52 |
| mikiobraun | ok, bye | 15:52 |
| @sonney2k | f-x, please don't forget to write the weekly email :D | 15:52 |
| @bettyboo | ha ha | 15:52 |
| -!- mikiobraun [~mikiobrau@squid.ml.tu-berlin.de] has left #shogun [] | 15:52 | |
| @sonney2k | and it never hurts to idle in dthis channel - someone is usually here to help | 15:52 |
| f-x | sonney2k: i won't! :) | 15:53 |
| @sonney2k | f-x, thanks! | 15:53 |
| f-x | sonney2k: and do i need to mail for this week? | 15:53 |
| f-x | because i mailed one 3-4 days back | 15:53 |
| @sonney2k | heiko1, for your mid-term eval I think when you manage to get working what you planned this week plus maybe that from the python_modular side is totally sufficient to me | 15:54 |
| heiko1 | ok alright | 15:54 |
| heiko1 | think this should be doable | 15:54 |
| heiko1 | I think the part with the parameter combinations was the hardest part already | 15:55 |
| heiko1 | now it comes to putting things together | 15:55 |
| heiko1 | ne subset stuff for the remaining classes is a bigger thing, but besides from that, some basic model selection should work till mid-term | 15:55 |
| heiko1 | sorry, 5min afk | 15:56 |
| blackburn | see you | 15:56 |
| -!- blackburn [5f43cffd@gateway/web/freenode/ip.95.67.207.253] has quit [Quit: Page closed] | 15:56 | |
| @sonney2k | f-x, end of this week is totally sufficient | 15:57 |
| f-x | sonney2k: okay. i'll mail for next week.. | 15:57 |
| @sonney2k | f-x, about the buffering fixed size vs. fixed number of examples | 15:59 |
| f-x | yes, sonney2k | 15:59 |
| f-x | john says it's more convenient to use fixed examples | 15:59 |
| @sonney2k | What would you like to implement? | 15:59 |
| @sonney2k | He has more experience in this | 16:00 |
| f-x | fixed examples seems easier | 16:00 |
| @sonney2k | at least he implemented it in vw and it works :D | 16:00 |
| f-x | very well too :) | 16:00 |
| f-x | fixed buffer size is also possible, but i at least need a decent approximation for the initial ring size in terms of examples | 16:00 |
| @sonney2k | so I would always do the proving thing first | 16:01 |
| @sonney2k | it could be problematic if you have some examples that are huge and you just didn't expect that | 16:01 |
| f-x | sorry, didn't get it exactly | 16:02 |
| @sonney2k | if you have a buffer for say 4 examples and suddenly they all are 1G or so | 16:02 |
| @sonney2k | you might run out of memory | 16:02 |
| @sonney2k | this wouldn't happen if you have a fixed buffer | 16:02 |
| @sonney2k | (this has other problems though - if examples are too big you cannot load them) | 16:03 |
| f-x | even with a fixed buffer, it may be inefficient unless we allow it to expand dynamically | 16:03 |
| f-x | how does the user know what size of the buffer to set? | 16:03 |
| heiko1 | back | 16:04 |
| @sonney2k | f-x, yes that is the problem of using a fixed buffer. | 16:05 |
| f-x | but your concern is valid, sonney2k.. vw doubles the buffer size if insufficient | 16:06 |
| f-x | and if you try to store 4 complete examples, 1G each in the buffer, memory problems guaranteed | 16:06 |
| f-x | i need to see how the vw code handles that | 16:06 |
| f-x | or consult John | 16:06 |
| @sonney2k | f-x, anyway I would say you implement the example based setting - we could set a maximum buffer size later on but for now it is the way to start | 16:06 |
| f-x | sonney2k: okay. we could always change it later | 16:07 |
| f-x | it wouldn't change many parts of the code, only a couple of functions | 16:07 |
| f-x | sonney2k: one more question | 16:08 |
| f-x | do you think it is better if the framework is first done completely, and then i begin with the vw integration | 16:08 |
| @sonney2k | f-x, I am thinking about a midterm goal and maybe it would be a good idea to at least have for simple / sparse features of any type the streaming features ready with some example showing how it works | 16:08 |
| f-x | or should i do it side by side | 16:08 |
| @sonney2k | or integrate that with e.g. SGD or so and not do all features | 16:09 |
| f-x | sonney2k: gtg, back in 10 minutes | 16:09 |
| f-x | sorry | 16:09 |
| @sonney2k | ok cu | 16:09 |
| @sonney2k | finally! | 16:09 |
| @sonney2k | thunderstorm | 16:09 |
| @sonney2k | 30 C here... | 16:09 |
| heiko1 | hehe :) | 16:09 |
| heiko1 | same here | 16:09 |
| heiko1 | but no storm yet | 16:10 |
| @sonney2k | drop temperature drop! | 16:10 |
| heiko1 | but will come in ~2 hrs :) | 16:10 |
| @sonney2k | heiko1, good luck :) | 16:10 |
| heiko1 | just survived a thunderstorm in a tent in france some days ago :) | 16:10 |
| @sonney2k | I did yesterday swimming in a lake | 16:11 |
| heiko1 | nice :) | 16:11 |
| heiko1 | which one? | 16:11 |
| @sonney2k | as long as one can see the shore it is all good I suppose | 16:11 |
| @sonney2k | don't know the name close to Fürstenwalde | 16:12 |
| heiko1 | ok, don't know | 16:12 |
| @sonney2k | http://maps.google.de/maps?f=q&source=s_q&hl=en&geocode=&q=f%C3%BCrstenwalde&aq=&sll=51.151786,10.415039&sspn=16.314769,46.538086&ie=UTF8&hq=&hnear=F%C3%BCrstenwalde,+Oder-Spree,+Brandenburg&ll=52.399486,14.03821&spn=0.030951,0.090895&t=h&z=14 | 16:12 |
| @sonney2k | trebuser see | 16:12 |
| heiko1 | hehe hail google maps .) | 16:14 |
| @sonney2k | yeah they have some nice programs | 16:15 |
| heiko1 | yes, indeed | 16:15 |
| heiko1 | also the gsoc programe is really cool | 16:15 |
| @sonney2k | heiko1, heh :) | 16:16 |
| * sonney2k agrees | 16:16 | |
| heiko1 | in a few years, they'll own the world, I am sure. | 16:17 |
| @sonney2k | heiko1, anyway I think you made good progress. There will be lots of seemingly minor side issues I suspect in the end | 16:17 |
| heiko1 | thanks :) | 16:17 |
| heiko1 | yes, I think so | 16:17 |
| heiko1 | many small things | 16:17 |
| @sonney2k | like the SVM C thing | 16:17 |
| heiko1 | yes | 16:18 |
| @sonney2k | I didn't have time to deal with that yet btw | 16:18 |
| heiko1 | In the example, I now just changed the C1 parameter | 16:18 |
| heiko1 | I think this can be done some time later | 16:18 |
| @sonney2k | but when you make progress like this we will see the many problems soon | 16:19 |
| heiko1 | still dont know which I prefer, hacking around in the CSVM class or making it possible to have multiple paramters under one name | 16:19 |
| f-x | sonney2k: back | 16:21 |
| f-x | i'm starting to think it is better if i start with some learning algorithm simultaneously | 16:21 |
| f-x | will probably help to iron out some problems with the streaming features backend | 16:21 |
| f-x | as you say, SGD is a good option | 16:21 |
| @sonney2k | heiko1, I mean it is confusing to have 2 C's but wanting to specify only one of the C's | 16:22 |
| heiko1 | what do you mean? | 16:23 |
| @sonney2k | heiko1, I would say shogun has no notion of SVM-C | 16:23 |
| @sonney2k | it has two C's | 16:23 |
| heiko1 | ah ok | 16:23 |
| @sonney2k | and that creates the confusion | 16:23 |
| @sonney2k | f-x, yes good idea | 16:24 |
| @sonney2k | f-x, I am concerned that the we cannot as originally planned derive e.g. SimpleFeatures from StreamingFeatures or so | 16:24 |
| f-x | sonney2k: hmm.. i thought that might be a problem | 16:25 |
| f-x | or do you want me to change the code in some way that it is possible it derive it? | 16:25 |
| @sonney2k | f-x, the problem is templates | 16:26 |
| @sonney2k | one cannot derive from a templated class and still be general | 16:26 |
| @sonney2k | I mean type T has to be defined then | 16:27 |
| f-x | oh.. i get it | 16:27 |
| f-x | so StreamingFeatures has to be an independent templated class? | 16:27 |
| @sonney2k | I don't know yet | 16:28 |
| f-x | right now the line reader has only been implemented for float64_t | 16:28 |
| @sonney2k | f-x, it would be great if it was compatible to CDotFeatures http://www.shogun-toolbox.org/doc/classshogun_1_1CDotFeatures.html | 16:29 |
| @sonney2k | because that is what the linear SVMs currently use | 16:29 |
| @sonney2k | f-x, if we just do the streamingfeatures separately it is easy | 16:29 |
| @sonney2k | then we could template and have the online algorithm use it | 16:30 |
| @sonney2k | it would be great to interact with the batch based features that we have now though. | 16:31 |
| @sonney2k | I don't know if this is easily possible | 16:31 |
| @sonney2k | hmmhh | 16:31 |
| @sonney2k | maybe one could derive from CDotFeatures - a CStreamingDotFeatures class that takes a StreamingFeatures object | 16:32 |
| @sonney2k | and same for the Simple/Sparse/StringFeatures | 16:32 |
| f-x | yes, seems possible | 16:32 |
| @sonney2k | or alternatively hack this direclty into these classes | 16:33 |
| f-x | i haven't gone through all the code for DotFeatures yet, but i'm thinking many functions will have to be redefined | 16:33 |
| f-x | in the derived class | 16:33 |
| @sonney2k | basically only what we have in SimpleFeatures | 16:33 |
| @sonney2k | same code actually - but it is weird | 16:34 |
| @sonney2k | I think we should start separately | 16:35 |
| f-x | start separately? | 16:35 |
| f-x | like we have done so far? | 16:35 |
| @sonney2k | then when the code manifests - after midterm eval figure out what we could generalize | 16:35 |
| @sonney2k | yes | 16:35 |
| @sonney2k | f-x, so I think you should define an OnlineMachine - in spirit of CLinearMachine for example | 16:36 |
| @sonney2k | and then derive SGD or liblinear from that | 16:37 |
| f-x | sonney2k: yes.. and should i start work on that now? | 16:37 |
| @sonney2k | that means code duplication though | 16:37 |
| @sonney2k | f-x, well first finish the buffer stuff and add some small example code that illustrates how to use the streaming features | 16:37 |
| @sonney2k | we can add a toy data set for that too | 16:38 |
| f-x | okay, but it will only work for simple input formats now | 16:38 |
| @sonney2k | I think you also need to do some benchmarks how fast the stuff is | 16:38 |
| @sonney2k | yes I know | 16:38 |
| @sonney2k | you need to make it more general I am afraid such that we can have a drop in replacement with a binary format later | 16:39 |
| f-x | it won't be much of a problem, just defining separate line reading functions | 16:40 |
| @sonney2k | I mean you currently parse a line -> example. Now same thing for other stuff | 16:40 |
| @sonney2k | yes | 16:40 |
| @sonney2k | it shouldn't be difficult | 16:40 |
| f-x | it's just that i need to know the exact definitions of the input formats | 16:40 |
| @sonney2k | btw we need a line parser for that too | 16:41 |
| @sonney2k | getline :D | 16:41 |
| f-x | sonney2k: i can do that, but was thinking of dropping getline altogether | 16:42 |
| f-x | but i think for now getline is okay | 16:42 |
| f-x | i'll submit it | 16:42 |
| f-x | :) | 16:42 |
| @sonney2k | f-x, yes for your code you need to drop it - but we need that in the other ascii based features too | 16:43 |
| @sonney2k | so it is definitely not wasted time | 16:43 |
| f-x | sonney2k: sure.. i'll make one using fread | 16:43 |
| @sonney2k | f-x, or fgets - just check if that is fast | 16:43 |
| @sonney2k | (enough) | 16:43 |
| f-x | okay | 16:44 |
| f-x | so getline is first priority now | 16:44 |
| @sonney2k | f-x, so assume that this is done | 16:44 |
| f-x | hmm | 16:44 |
| @sonney2k | the current CSGD code relies on CDotFeatures as inputs | 16:44 |
| @sonney2k | same for liblinear | 16:45 |
| f-x | so i'm afraid the StreamingDotFeatures will have to be done right now | 16:45 |
| @sonney2k | The problem is that these algorithms are working in batch mode currenlty | 16:46 |
| @sonney2k | i.e. they do multiple passes over the data | 16:46 |
| f-x | yeah | 16:46 |
| @sonney2k | one can set them to epochs=1 though | 16:46 |
| f-x | so you want a flag to specify online/offline mode? | 16:47 |
| @sonney2k | and also IIRC they all currenlty require that the number of examples is known beforehand | 16:47 |
| f-x | i mean, the online one shouldn't be a separate file? | 16:47 |
| @sonney2k | if we don't want too much code duplication we need to have such a flag | 16:47 |
| heiko1 | sonney2k, is it possible to define abstract base classes in shogun? because class_list.cpp complains, if I do. | 16:48 |
| @sonney2k | f-x, I guess you will have to look at CDotFeatures and then try to derive things | 16:49 |
| @sonney2k | when problems occur ask | 16:50 |
| @sonney2k | but first getline and example etc | 16:50 |
| f-x | hmm.. correct. only then can i begin with an online algorithm | 16:50 |
| @sonney2k | yes | 16:50 |
| @sonney2k | heiko1, then they shouldn't appear in class_list* | 16:50 |
| heiko1 | how do i do that? not extend CSGObject? | 16:51 |
| @sonney2k | heiko1, use https://gist.github.com/ to show me te code | 16:51 |
| f-x | so i'll get to the getline thing first | 16:51 |
| f-x | see you guys later, sonney2k and heiko1 | 16:51 |
| heiko1 | bye :) | 16:51 |
| @sonney2k | f-x, yeah cu! | 16:51 |
| @sonney2k | and be around more often :) | 16:52 |
| heiko1 | https://gist.github.com/1010390 | 16:53 |
| heiko1 | something like this | 16:53 |
| @sonney2k | but don't use SGVector& | 16:54 |
| @sonney2k | we want values copied | 16:55 |
| @sonney2k | it should work though | 16:55 |
| @sonney2k | (not appear in class_list*) | 16:55 |
| heiko1 | mmh ... | 16:55 |
| heiko1 | it does, ill do a make clean | 16:55 |
| @sonney2k | or git clean -dfx | 16:56 |
| @sonney2k | (beware that this ereases all files not in git) | 16:56 |
| heiko1 | thanks for the warning :) | 16:57 |
| @bettyboo | ;) | 16:57 |
| heiko1 | what makes a class appear in class_list*? | 16:57 |
| @sonney2k | it starts with a C prefix and has no pure virtual functions | 16:58 |
| @sonney2k | look at my class_list.cpp.py script for the magic :) | 16:58 |
| heiko1 | ok :) | 16:59 |
| heiko1 | ok works now, thanks | 17:02 |
| * heiko1 will get a coffee | 17:03 | |
| @sonney2k | gtg | 17:07 |
| @sonney2k | l8r | 17:07 |
| -!- blackburn [~blackburn@188.122.237.128] has joined #shogun | 17:23 | |
| -!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 17:30 | |
| -!- f-x [~shashwat@117.192.199.159] has quit [Ping timeout: 276 seconds] | 17:32 | |
| -!- blackburn [~blackburn@188.122.237.128] has quit [Quit: Leaving.] | 17:49 | |
| -!- heiko1 [~heiko@134.91.10.201] has left #shogun [] | 18:12 | |
| -!- f-x [~shashwat@117.192.207.225] has joined #shogun | 19:09 | |
| -!- f-x [~shashwat@117.192.207.225] has quit [Client Quit] | 19:10 | |
| @sonney2k | Chuck Norris has only one OSI level - Physical | 20:59 |
| @sonney2k | great stuff | 20:59 |
| CIA-18 | shogun: Soeren Sonnenburg master * r54230de / src/python_modular/swig_typemaps.i : Support all datatypes from SGDataType in new python_modular typemaps. - http://bit.ly/lR5HT5 | 22:16 |
| -!- serialhex_ [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 23:15 | |
| -!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Ping timeout: 255 seconds] | 23:16 | |
| -!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 23:42 | |
| alesis-novik | Hello. Sorry I missed the meeting, but I had prior arrangements from before the meeting time was announced. | 23:43 |
| alesis-novik | I'll go over the logs | 23:43 |
| --- Log closed Tue Jun 07 00:00:04 2011 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!