--- Log opened Sun Dec 29 00:00:27 2013 | ||
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 00:02 | |
lisitsyn | HeikoS: hey there | 00:05 |
---|---|---|
@HeikoS | lisitsyn: hi! | 00:05 |
lisitsyn | HeikoS: how are you? having holidays? | 00:06 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 00:06 | |
shogun-notifier- | shogun: Saurabh7 :develop * cdc5a1b / examples/undocumented/libshogun/evaluation_cross_validation_knn.cpp: https://github.com/shogun-toolbox/shogun/commit/cdc5a1b23ca0302e74960afc50c97c2f1aea65a7 | 00:06 |
shogun-notifier- | shogun: knn example | 00:06 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 7d2ad10 / examples/undocumented/libshogun/evaluation_cross_validation_knn.cpp: https://github.com/shogun-toolbox/shogun/commit/7d2ad10b8228eff00a4d966a978f7974df9006ab | 00:06 |
shogun-notifier- | shogun: Merge pull request #1797 from Saurabh7/xvalexamples | 00:06 |
shogun-notifier- | shogun: | 00:06 |
shogun-notifier- | shogun: knn example | 00:06 |
@HeikoS | lisitsyn: yeah more or less | 00:06 |
@HeikoS | lisitsyn: not really, exams soon | 00:06 |
lisitsyn | uh | 00:06 |
@HeikoS | but taking it relaxed | 00:06 |
@HeikoS | working less than I should, getting stressed more than I should ;) | 00:06 |
@HeikoS | lisitsyn: how about yourself? | 00:07 |
@sonney2k | HeikoS, have you seen we have Chiyuan's notebooks now | 00:08 |
@HeikoS | sonney2k: hi! yes I saw that | 00:08 |
@HeikoS | sonney2k: I also pushed one half-finished on X-validation | 00:08 |
@sonney2k | nice | 00:10 |
@HeikoS | sonney2k: pretty nice the multiclass ones | 00:10 |
@HeikoS | sonney2k: just reading besser82 's proposal | 00:10 |
lisitsyn | HeikoS: finishing stuff :) | 00:10 |
@sonney2k | HeikoS, I wanted to do some evaluation notebook soon | 00:11 |
@sonney2k | like ROC/ PRC and all the measures | 00:11 |
@HeikoS | sonney2k: yeah nice | 00:11 |
@HeikoS | sonney2k: there is some of this in the x-validation notebook | 00:11 |
@sonney2k | yeah just seen | 00:11 |
@sonney2k | but not really overlapping | 00:11 |
@HeikoS | I agree | 00:11 |
@HeikoS | its good to have another one | 00:11 |
@sonney2k | HeikoS, we should maybe email all ex-gsoc / contributors asking if they can do a notebook for their work | 00:12 |
@sonney2k | like lisitsyn :P | 00:12 |
@HeikoS | sonney2k: haha ;) | 00:12 |
@HeikoS | yeah lisitsyn should totally do one | 00:12 |
@sonney2k | otherwise work is pretty lost | 00:12 |
lisitsyn | yeah I will | 00:12 |
@sonney2k | just too deep hidden inside shogun | 00:13 |
@HeikoS | sonney2k: I know, but we have already quite a few | 00:13 |
lisitsyn | we just have to believe and I'll do that | 00:13 |
lisitsyn | :D | 00:13 |
@HeikoS | sonney2k: could also work on making things easier accessible ;) | 00:13 |
@sonney2k | HeikoS, ? | 00:13 |
@HeikoS | sonney2k: like a feature/example list that shows very clearly what is where ... but that is not really concrete thoughts | 00:14 |
@HeikoS | sonney2k: did you end up pushing the release more? | 00:14 |
@sonney2k | seems like besser82 won't finish his cmake stuff so I guess we can release 3.1 | 00:14 |
@sonney2k | well it is in OK state | 00:14 |
@sonney2k | we could release any time that means | 00:14 |
@sonney2k | well news needs an update and we have to think were we announce it | 00:15 |
@sonney2k | but good enough for sure | 00:15 |
@sonney2k | HeikoS, what I have in mind with notebooks is that we at some point put in the table of contents that you had in the original tutorial | 00:15 |
@sonney2k | so it will kind of become a book | 00:16 |
@sonney2k | no idea if you have seen stuff like machine learning for hackers etc | 00:16 |
@sonney2k | I am just sure we can do better by doing it this way | 00:16 |
@HeikoS | sonney2k: I agree! | 00:16 |
@HeikoS | sonney2k: cool about the release | 00:17 |
@sonney2k | HeikoS, about the plugin stuff | 00:17 |
@sonney2k | it is still not thought through | 00:17 |
@HeikoS | sonney2k: yeah notebooks can be presented in a better way, more centralised somehow, but for now the current way is also fine. I think this will become obsolete thouigh when the number is growing | 00:17 |
@HeikoS | sonney2k: yeah, just sent a mail, this needs ultra careful planning | 00:18 |
@sonney2k | but yes it would be nice if it would work | 00:18 |
@HeikoS | and breath | 00:18 |
@sonney2k | yeah just read it | 00:18 |
@HeikoS | and maybe a gsoc student | 00:18 |
@sonney2k | no | 00:18 |
@sonney2k | gsoc student won't be enough for that | 00:18 |
@HeikoS | sonney2k: do you think thats too boring? | 00:18 |
@HeikoS | sonney2k: more as a helping hand | 00:18 |
@sonney2k | no no | 00:19 |
@HeikoS | (embedded into a project) | 00:19 |
@sonney2k | it could just be some proof of concept | 00:19 |
@HeikoS | sonney2k: btw I got my new Thinkpad t440s | 00:19 |
@HeikoS | thats quite nice | 00:19 |
@HeikoS | sonney2k: what are you thoughts in the licensing btw? | 00:20 |
@sonney2k | HeikoS, very difficult | 00:20 |
@sonney2k | one thing is pretty clear to me | 00:20 |
@sonney2k | if we change (are allowed to / manage) | 00:21 |
@sonney2k | we should change to BSD / MIT | 00:21 |
@HeikoS | sonney2k: question is if it is worth it | 00:21 |
@HeikoS | sonney2k: yeah I agree | 00:21 |
@sonney2k | LGPL is no different to GPL for companies | 00:21 |
@sonney2k | but I don't know if it is worth it | 00:21 |
@HeikoS | I think so | 00:21 |
@sonney2k | and even possible | 00:21 |
@sonney2k | I mean there is a reason I never wrote (C) Soeren Sonnenburg | 00:21 |
@HeikoS | sonney2k: I think this would be a longer term project | 00:21 |
@sonney2k | all my work was owned by Fraunhofer | 00:22 |
@HeikoS | ah | 00:22 |
@HeikoS | sonney2k: I think we should have alist of such things that might cause problems with that | 00:22 |
@HeikoS | and then maybe write all authors we can identify, and then see | 00:23 |
@HeikoS | first get some better feeling for feasibility and then decide | 00:23 |
@sonney2k | and the only reason so far to attempt that would be if libshogun would become kind of the c/c++ backend of scikits-learn | 00:23 |
@sonney2k | or so | 00:23 |
@HeikoS | sonney2k: maybe | 00:23 |
@HeikoS | sonney2k: I am thinking more of talking to companies to get them interested in the project | 00:23 |
@HeikoS | and to avoid this "no gpl" barrier that many people have (which is wring I think, but thats how it is) | 00:24 |
@HeikoS | sonney2k: man you should have been at the mloss workshop, quite strong feelings against GPL by many people | 00:24 |
@sonney2k | yeah companies... I personally don't really want M$ / apple to just use shogun | 00:25 |
@sonney2k | I mean embedded | 00:25 |
@sonney2k | they can train their stuff sure | 00:25 |
@sonney2k | but not inside the M$ 'some big-data product name' | 00:25 |
@sonney2k | HeikoS, no idea if an evaluation BSD module or so would help companies | 00:26 |
@sonney2k | this IMHO would be something we could do | 00:26 |
@HeikoS | sonney2k: I think any usage would be good for us | 00:27 |
@HeikoS | sonney2k: but yeah evaluation module is also an idea | 00:27 |
lisitsyn | I find it very optimistic to expect big companies use shogun ;) | 00:28 |
@sonney2k | HeikoS, well GPL doesn't restrict any usage except for embedding in a product to be sold right? | 00:28 |
@HeikoS | sonney2k: if shogun would be used in such a context, it might benefit us since getting feedback, or even some kind of support. If not thats also fine, since currently we dont get this. | 00:28 |
@sonney2k | lisitsyn, haha | 00:28 |
@HeikoS | sonney2k: I know, but you know that people sometimes think different | 00:28 |
@sonney2k | lisitsyn, you are living under a rock then | 00:28 |
@HeikoS | and try to avoid any GPL stuff | 00:28 |
lisitsyn | sonney2k: I just never had any proofs ;) | 00:29 |
@sonney2k | lisitsyn, there is a reason why shogun-toolbox.org gets more hits mon-fri's | 00:29 |
@sonney2k | HeikoS, yes true some companies do that | 00:30 |
lisitsyn | is that the fact supporting your hypothesis? | 00:30 |
lisitsyn | ;) | 00:30 |
@sonney2k | lisitsyn, no - but people telling me / companies asking | 00:30 |
lisitsyn | that sounds better | 00:30 |
@HeikoS | sonney2k: my point is more to make usage as easy as possible, to allow for more interaction with whomever | 00:31 |
@sonney2k | but I certainly for any basic stuff people use sckits-learn | 00:31 |
@sonney2k | HeikoS, yes sure I get that | 00:32 |
lisitsyn | recently I had to show some basic classification | 00:32 |
lisitsyn | it took 4 lines in scikits | 00:32 |
lisitsyn | with hashing vectorizer and svm | 00:32 |
lisitsyn | if we are not down to just a few lines for such tasks we lose | 00:32 |
@sonney2k | no idea what hashing vectorizer is | 00:33 |
@sonney2k | but shogun syntax is as compact | 00:33 |
@HeikoS | lisitsyn, sonney2k what do you think about joining forces with scikit a bit? I talked to a few of them and there was the idea to have a joint code sprint for example | 00:33 |
@HeikoS | lisitsyn: yeah we should not be afraid of them | 00:33 |
@sonney2k | HeikoS, that would only work if licenses were compatible | 00:33 |
lisitsyn | HeikoS: I don't know how can we join together | 00:33 |
@HeikoS | sonney2k: yep | 00:33 |
@HeikoS | lisitsyn: maybe not code but talking | 00:33 |
@sonney2k | HeikoS, that is why I said above the only incentive I see is being their backend | 00:33 |
lisitsyn | I see | 00:34 |
@HeikoS | you know, how do you solve XYZ ? | 00:34 |
lisitsyn | they don't need backend I think | 00:34 |
@HeikoS | sonney2k: I dont believe that this will happen, but might be an idea | 00:34 |
@HeikoS | sonney2k: like have a shogun wrapper in scikit | 00:34 |
lisitsyn | I think scikits should be the backend of shogun | 00:34 |
@HeikoS | or so ... | 00:34 |
lisitsyn | :D | 00:34 |
lisitsyn | I will continue pushing the idea of some kind of integration platform | 00:34 |
@sonney2k | HeikoS, yeah but just relicensing for some potential companies that want to sell big data analysis toolkits as closed source software? | 00:34 |
@sonney2k | not really worth the effort IMHO | 00:35 |
@HeikoS | sonney2k: happened to ipython and now they got a massive dontation from MS | 00:35 |
@HeikoS | donation | 00:35 |
lisitsyn | our pov differs here | 00:35 |
@HeikoS | plus lots of code given back | 00:35 |
lisitsyn | I guess me and HeikoS on the same side with 'more usage is better' | 00:35 |
@HeikoS | exactly, I dont like that happening without getting anything back too | 00:36 |
@HeikoS | but I think that there always comes something back | 00:36 |
@HeikoS | also would be cool for us to know that people use it | 00:36 |
lisitsyn | you don't get back if you require it | 00:36 |
@HeikoS | more users attract more users | 00:36 |
@sonney2k | I don't like writing code that M$ or other companies can just 'steal' | 00:36 |
lisitsyn | it makes more sense to have a few badges 'used by X' | 00:36 |
lisitsyn | than to have nothing and be proud of freedom ;) | 00:37 |
@sonney2k | depends on what you want to do right? | 00:37 |
@HeikoS | sonney2k: totally, thats why we are discussing this right? | 00:37 |
@sonney2k | I wrote the algorithms I and later others needed | 00:38 |
@HeikoS | but its certainly not top priority | 00:38 |
lisitsyn | well I speak for my code but I'd be happy if anyone used it | 00:38 |
@HeikoS | I would rather do this slowly in the background step by step | 00:38 |
lisitsyn | I see no difference between getting no money because of no usage and getting no money because of 'stealing' | 00:39 |
@sonney2k | lisitsyn, what would we do with money and shogun? | 00:39 |
@HeikoS | its not really about money, more about feedback | 00:39 |
lisitsyn | sonney2k: nothing I just don't know what is expected outcome of using software in companies | 00:39 |
lisitsyn | if you restrict them so much they avoid using it you get no feedback | 00:40 |
lisitsyn | liblinear is in any iphone | 00:40 |
lisitsyn | I see that kind of success | 00:40 |
@sonney2k | lisitsyn, well any sane person would use shogun if you not on a anti GPL trip | 00:40 |
@sonney2k | lisitsyn, and if $COMPANY uses this for whatever *I* would want it to use it under the GPL terms | 00:43 |
lisitsyn | sonney2k: yeah I get you want to keep gpl here :) | 00:44 |
@sonney2k | lisitsyn, I don't see any benefit of converting to BSD so far | 00:46 |
@sonney2k | lisitsyn, liblinear is just a tiny algorithm | 00:46 |
lisitsyn | sonney2k: I can't say I am a big fan of converting stuff too | 00:47 |
lisitsyn | it sounds like useless spending of time though I believe it should have been done under bsd | 00:48 |
@sonney2k | the way I see it - shogun is a big framework for doing experiments | 00:48 |
@sonney2k | once you know which algorithm you use | 00:48 |
@HeikoS | sonney2k, lisitsyn this is all not really about the license consequences, but more about how other people see the license. | 00:49 |
@sonney2k | you can as well find some bsd / whatever license you need algorithm or write that little bit | 00:49 |
@sonney2k | and use it in your commercial product that you want to ship | 00:50 |
@sonney2k | but most people don't ship stuff | 00:50 |
@sonney2k | they create some prototype | 00:50 |
@sonney2k | and that is what shogun is made for | 00:50 |
@sonney2k | it is totally different from the liblinear is in iphone thing | 00:50 |
@sonney2k | HeikoS, yeah - but heck people compile with gcc for decades and did never complain that it produces executables that are even commerically shipped | 00:51 |
@sonney2k | I see us to be the very same thing | 00:52 |
@sonney2k | the end result is some model | 00:52 |
@sonney2k | that applying is usually ultra trivial | 00:52 |
@sonney2k | and the model itself is *not* GPL - not at all | 00:52 |
@sonney2k | it is whatever you want it to be | 00:52 |
lisitsyn | well actually the problem is not in license but market share | 00:55 |
-!- travis-ci [~travis-ci@ec2-23-22-142-220.compute-1.amazonaws.com] has joined #shogun | 00:55 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/16085444 | 00:55 |
-!- travis-ci [~travis-ci@ec2-23-22-142-220.compute-1.amazonaws.com] has left #shogun [] | 00:55 | |
lisitsyn | if you google for machine learning you get scikits | 00:55 |
lisitsyn | I am quite demotivated to work on anything that is not trying to be the best | 00:55 |
lisitsyn | HeikoS had a conversation with basically our customers | 00:56 |
lisitsyn | and they want bsd so that's why it started :) | 00:56 |
@HeikoS | lisitsyn: not really customers | 00:56 |
@HeikoS | its more fellows | 00:57 |
@HeikoS | that shared their experiences | 00:57 |
lisitsyn | HeikoS: well anyway people who could used it | 00:57 |
@HeikoS | lisitsyn: Its more the mood in the discussion that made me clear that GPL is a barrier, which I wasnt aware of before | 00:57 |
lisitsyn | HeikoS: yeah it is quite trendy to avoid gpl nowadays | 00:58 |
lisitsyn | I feel it started like a few years back | 00:58 |
@HeikoS | lisitsyn: I actually agree with it, but not strongly and emotionally ;) | 00:59 |
@sonney2k | lisitsyn, scikits has high market share for other reasons - they cover the basic algorithms and make things easily accessible | 01:00 |
@sonney2k | lisitsyn, totally the opposite of what shogun was back then - state-of-the-art expert-only algorithms | 01:01 |
@HeikoS | I would like to see shogun more towards the basic stuff btw | 01:03 |
lisitsyn | we have no resources but heiko to write advanced stuff | 01:04 |
lisitsyn | I am not and not going to be a researcher I think | 01:04 |
@HeikoS | lisitsyn: then you should do the basic framework stuff | 01:04 |
@HeikoS | we need that desperately | 01:05 |
@HeikoS | I can/will add more fundamental things | 01:05 |
@HeikoS | that attract users | 01:05 |
@HeikoS | loads of ideas, just not enough time | 01:05 |
lisitsyn | well simply it is | 01:05 |
@sonney2k | HeikoS, I also want basics you know that :) | 01:05 |
@HeikoS | sonney2k: I know, we talked about that :) | 01:05 |
lisitsyn | any algorithm should look like 50-100 lines of code | 01:05 |
@sonney2k | HeikoS, at some point please scan in the notes you took | 01:05 |
@HeikoS | sonney2k: will do | 01:05 |
lisitsyn | if it is not we've got wrong api | 01:06 |
@HeikoS | lisitsyn: the interface is quite nice, (with exceptions) | 01:06 |
@HeikoS | and we added all these tests which was a great step forward | 01:06 |
@HeikoS | I guess next thing should be presentation | 01:06 |
@HeikoS | and then clean ups of core | 01:06 |
lisitsyn | yeah may be | 01:07 |
@HeikoS | while adding more basic algorithms (we definitely do not need more fancy complicated ones for now=) | 01:07 |
@HeikoS | lisitsyn: I would love to decouple the parameter framework for example like your stichwort stuff | 01:07 |
@HeikoS | that would be cool | 01:07 |
lisitsyn | HeikoS: I am disappointed about its usage | 01:07 |
lisitsyn | I tried it and I don't feel comfortable | 01:08 |
lisitsyn | something should be changed | 01:08 |
@HeikoS | maybe host a GSoC project on that | 01:08 |
@HeikoS | wiuth the goal to replace the current system | 01:08 |
@HeikoS | would be interesting proper c++ coding stuff | 01:08 |
lisitsyn | if I can't come with some advanced C++ stuff student could have more troubles ;) | 01:08 |
shogun-buildbot | build #152 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/152 blamelist: Saurabh7 <saurabh.mahindre@gmail.com> | 01:09 |
@HeikoS | lisitsyn: but if you design a thing and somebody does it for you? | 01:09 |
lisitsyn | once it is designed it is piece of cake | 01:09 |
@sonney2k | what is not? | 01:09 |
@sonney2k | but still it needs manpower to implement | 01:09 |
lisitsyn | HeikoS: I am still trying to find a balance between compile time stuff and runtime stuff | 01:09 |
@HeikoS | lisitsyn: design is also ongoing, you write down stuff, student implements, realises there are flaws, you update design | 01:10 |
lisitsyn | well why not | 01:10 |
@HeikoS | lisitsyn: I did that with lambday this way and it worked great (for parallel framework and log-det stuff) | 01:10 |
lisitsyn | I just mean if we set up task 'design that' | 01:10 |
@sonney2k | HeikoS, btw we need to be rather quick if we want to be in next years gsoc | 01:10 |
lisitsyn | he could get mad :D | 01:10 |
@sonney2k | we need that list of ideas and find mentors | 01:10 |
lisitsyn | okay guys see you later | 01:11 |
@sonney2k | HeikoS, I know we have the ideas :) | 01:11 |
@HeikoS | sonney2k: I know, I started writing some things down already (personal drafts) and talked to a few possible additional mentors | 01:11 |
@sonney2k | but mentors? | 01:11 |
@HeikoS | some people in my institute might want to mentor things | 01:11 |
@HeikoS | have one for neural networks, one for probabilistic learning, and maybe more | 01:12 |
@sonney2k | HeikoS, it is important to come up with a list of ideas (written down even drafty) rather soon | 01:12 |
@HeikoS | sonney2k: but I currently really have a hard time with my time ... | 01:12 |
@sonney2k | recall that we got all the good students long before gsoc was announced | 01:12 |
@HeikoS | sonney2k: how is it going with you? | 01:12 |
@sonney2k | bad bad | 01:12 |
@HeikoS | damn | 01:12 |
@HeikoS | yeah I will try to do something in some evenings | 01:13 |
@sonney2k | I can only do the release now before vacations are over | 01:13 |
@HeikoS | sonney2k: go for it, might help | 01:13 |
@sonney2k | HeikoS, it might be sufficient to write an email to the mailinglist | 01:13 |
@HeikoS | Ill try to copy our ideas list and extend it | 01:13 |
@sonney2k | like we want to participate in gsoc'14 | 01:13 |
@HeikoS | if I manage to do this before end of the year, I will send this attached with the mail. | 01:14 |
@HeikoS | if not, mail alone | 01:14 |
@HeikoS | lisitsyn: take care! | 01:14 |
@sonney2k | lisitsyn, sweet dreams :D | 01:14 |
@sonney2k | HeikoS, well pinging potential mentors is really important... | 01:15 |
@sonney2k | and that takes a *long* time | 01:16 |
@HeikoS | sonney2k: yeah, good that you remind me | 01:16 |
@HeikoS | will share a document with names ideas note etc | 01:16 |
@sonney2k | HeikoS, just FYI for last year mentors were already clear in November | 01:16 |
@HeikoS | uh | 01:17 |
@sonney2k | but we did add 1-2 IIRC later | 01:17 |
@HeikoS | sonney2k: well, yeah I cannot do magic. Will try to find as many as possible. It is also fine if we dont get 8 slots again | 01:17 |
@HeikoS | if we dont have the manpower to organise that, its no help anyways | 01:17 |
@sonney2k | yeah sure - I think I can get some too | 01:18 |
@sonney2k | you recall I want shogun & teaching | 01:18 |
@sonney2k | and I already talked to some people about this | 01:18 |
@HeikoS | I love the idea! | 01:21 |
shogun-buildbot | build #149 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/149 blamelist: Saurabh7 <saurabh.mahindre@gmail.com> | 01:25 |
shogun-buildbot | build #166 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/166 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Saurabh7 <saurabh.mahindre@gmail.com> | 02:11 |
shogun-buildbot | build #153 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/153 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 02:22 |
-!- HeikoS [~heiko@90.211.150.15] has left #shogun [] | 02:38 | |
shogun-buildbot | build #150 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/150 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 02:38 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 03:06 | |
shogun-buildbot | build #664 of nightly_default is complete: Failure [failed notebooks] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/664 | 04:53 |
-!- Saurabh7 [~Saurabh7@117.222.14.70] has joined #shogun | 06:32 | |
-!- Saurabh7 [~Saurabh7@117.222.14.70] has quit [Ping timeout: 245 seconds] | 07:06 | |
besser82 | sonney2k, ping? | 10:37 |
besser82 | sonney2k, http://www.luga.de/Aktionen/LIT-2014/CfP | 10:40 |
besser82 | sonney2k, they are looking for contribs for making presentations for stuff about "Why Linux?" | 10:40 |
besser82 | sonney2k, one sub-field of that is "Linux und Bildung" | 10:41 |
besser82 | sonney2k, like "Open-Source-Software in Schulen und Universit?ten zur Forschung" | 10:41 |
besser82 | sonney2k, how about doing a bit of lobby-work for SHOGUN?! | 10:42 |
besser82 | lisitsyn, Mornin' :D | 10:54 |
lisitsyn | besser82: MORGEN | 10:58 |
besser82 | lisitsyn, :D | 10:58 |
besser82 | lisitsyn, so how about plugins? | 10:59 |
lisitsyn | besser82: what plugins? | 10:59 |
besser82 | lisitsyn, do we want to setup a draft for those plans? | 11:00 |
lisitsyn | besser82: yeah probably | 11:00 |
@wiking | besser82: btw another question: would we still do dpointering | 11:00 |
besser82 | wiking, I think, when using plugins, they would be obsolete mostly | 11:01 |
@wiking | i mean in one sense it gets in a way deprecated by modularizing the whole thing... but then again having a stable ABI would be great that could be achieved by using dptrs | 11:01 |
besser82 | wiking, ABI / API would be stable, because the interface-lib wouldn't change that much | 11:02 |
besser82 | wiking, except for possibly adding some new class / datastructure / object | 11:02 |
besser82 | wiking, but that wouldn't break API / ABI | 11:03 |
besser82 | wiking, but might make sense to wrap the plugin-manager-stuff into d-ptrs | 11:04 |
@wiking | besser82: imho we should not have so fine grained modular arch that u suggested the other day | 11:05 |
@wiking | or at least not on repository level | 11:05 |
@wiking | besser82: maybe one repo per subdir of src/shogun/ | 11:07 |
besser82 | wiking, otoh, with fine grain we can be more flexible | 11:08 |
besser82 | wiking, but has overhead | 11:08 |
@wiking | besser82: yeah but that much repository | 11:08 |
@wiking | is just way too much | 11:08 |
besser82 | wiking, make this per subdir might be a bit complicated... | 11:08 |
@wiking | and waste of repo | 11:08 |
@wiking | besser82: it's like 1 file per repository | 11:08 |
@wiking | that's what you are suggesting | 11:08 |
@wiking | ok say 3 files: .h, .cpp and CMakeLists.txt | 11:09 |
besser82 | wiking, not really; I suggested sth. like logical grouping | 11:09 |
besser82 | wiking, you came up with "one plugin each feature" ;) | 11:09 |
@wiking | besser82: well 1 subdir per repo is a logical grouping | 11:10 |
@wiking | with a very few exception | 11:10 |
@wiking | "besser82: wiking: but if we are moving to plugin-style, I'd strongly recommend keeping every plugin inside a seperate git-repo" | 11:11 |
@wiking | from yesterday... | 11:11 |
@wiking | and 1 kernel implementation should be 1 plugin | 11:12 |
@wiking | that's why i infered -> 3 files per repo... | 11:12 |
besser82 | wiking, maybe we should do some logical grouping of repos | 11:13 |
@wiking | see my above suggestion | 11:13 |
@wiking | that is a logical grouping | 11:14 |
besser82 | wiking, lisitsyn probably wants to do some rubygem-style manager for those plugins | 11:27 |
besser82 | wiking, so that might be easier with more grained repos?!? | 11:27 |
besser82 | wiking, but in fact I'm open to anything | 11:28 |
besser82 | wiking, in the end sonney2k must approve all the crazy-shit, we want to do ;) | 11:29 |
besser82 | lisitsyn, do want to mock a spec about how to handle those plugins? | 11:30 |
lisitsyn | besser82: yes later | 11:30 |
lisitsyn | in the beginning of january | 11:30 |
besser82 | lisitsyn, you can ask ded moroz to bring one for you; so you just need to c&&p :-P | 11:32 |
lisitsyn | hah | 11:32 |
besser82 | wiking, so maybe you should discuss that repo-layout && stuff with lisitsyn | 11:34 |
besser82 | wiking, I would take mostly care about the code-pov, then | 11:35 |
@wiking | :> | 11:37 |
@wiking | yeah i think things will roll out as they usually do... | 11:39 |
@wiking | either something will happened or else | 11:39 |
@wiking | planning out who's going to do what is senseless in this stage | 11:40 |
@wiking | especially taking into consideration of all the other timeconstraints that eveyrbody has... | 11:40 |
@wiking | my 2 cents | 11:40 |
besser82 | wiking, maybe, maybe not | 11:40 |
@wiking | besser82: trustme it for sure like this | 11:40 |
@wiking | if u feel like doing it do it. put it out to a branch/fork whatever and some people will join if they can/feel like | 11:41 |
@wiking | i dont want to sound pessimitic or discouriging but this is how things have been in shogun ever since i'm around... | 11:41 |
@wiking | and that aint going to change as everybody is doing this parttime | 11:42 |
besser82 | wiking, lisitsyn came up the idea of making some specs first ;) | 11:43 |
@wiking | besser82: yes and that'll be done as well very sporadic | 11:44 |
@wiking | as it was in the past | 11:44 |
besser82 | wiking, from me pov I can just start, and we can do mods and building tools around that later ;) | 11:44 |
@wiking | there's going to be one googledoc/wiki page and people will start edit etc | 11:44 |
lisitsyn | haha | 11:44 |
lisitsyn | yes | 11:44 |
@wiking | besser82: is there anything that holds you back? | 11:44 |
@wiking | besser82: i mean if u can do it do it | 11:44 |
@wiking | eveyrbody is going to be happy about that | 11:44 |
besser82 | wiking, not having a spec && approval from sonney2k ? | 11:45 |
@wiking | ? | 11:45 |
@wiking | this is an open source project | 11:45 |
@wiking | u can do whatever you want with it | 11:45 |
@wiking | forkit and abuse it | 11:45 |
@wiking | if u want to do it alone that's fine as well if u want feedback make it out in the open, i.e. from the very beginning start writing the code in an open repo so ppl can comment | 11:46 |
@wiking | that's all | 11:46 |
@wiking | u dont need anybody's approval to do it | 11:46 |
besser82 | wiking, allright then ;) Lets-a-start ;) | 11:47 |
@wiking | "2) offer you this License | 11:48 |
@wiking | giving you legal permission to copy, distribute and/or modify it." | 11:48 |
@wiking | from gpl license | 11:48 |
@wiking | so that's all | 11:48 |
besser82 | wiking, lemme setup a fork and throw in my ideas to the wiki-page ;) | 11:48 |
@wiking | besser82: go ahead | 11:48 |
-!- new_lido [~walid@105.200.131.138] has joined #shogun | 15:05 | |
-!- iglesiasg [~iglesiasg@211.Red-83-40-129.dynamicIP.rima-tde.net] has joined #shogun | 15:26 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 15:26 | |
@iglesiasg | sup guys | 15:26 |
besser82 | iglesiasg, Hey! Hello! | 15:34 |
besser82 | iglesiasg, Read my proposal on da list? | 15:35 |
@iglesiasg | besser82, yes! However, I did not read the linked post yet | 15:38 |
@iglesiasg | so I have to read it first to understand better what changes would this approach involve | 15:39 |
besser82 | iglesiasg, kk | 15:45 |
besser82 | iglesiasg, Much work, real improvements ;) | 15:45 |
-!- new_lido [~walid@105.200.131.138] has quit [Read error: Connection reset by peer] | 15:48 | |
-!- new_lido [~walid@105.200.131.138] has joined #shogun | 15:48 | |
-!- new_lido [~walid@105.200.131.138] has quit [Quit: Leaving] | 17:33 | |
-!- iglesiasg [~iglesiasg@211.Red-83-40-129.dynamicIP.rima-tde.net] has quit [Ping timeout: 272 seconds] | 18:26 | |
-!- iglesiasg [~iglesiasg@211.Red-83-40-129.dynamicIP.rima-tde.net] has joined #shogun | 18:30 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 18:30 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 18:35 | |
shogun-notifier- | shogun: Saurabh7 :develop * b40ff33 / tests/unit/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/b40ff335612cc04f21e0365ff1e957db71bd001c | 18:35 |
shogun-notifier- | shogun: update splittingstrategy tests | 18:35 |
shogun-notifier- | shogun: Saurabh7 :develop * 7a4e760 / tests/unit/evaluation/SplittingStrategy_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/7a4e76074055fb6b78a3929b87201e80e132d9ac | 18:35 |
shogun-notifier- | shogun: update splittingstrategy tests | 18:35 |
shogun-notifier- | shogun: Heiko Strathmann :develop * cfcef10 / tests/unit/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/cfcef108b0f8a0f8cc8ecd2ff997b7e913d50a5e | 18:35 |
shogun-notifier- | shogun: Merge pull request #1802 from Saurabh7/xvaltests | 18:35 |
shogun-notifier- | shogun: | 18:35 |
shogun-notifier- | shogun: update Xvaltests | 18:35 |
-!- HeikoS [~heiko@0541999e.skybroadband.com] has joined #shogun | 18:45 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 18:45 | |
shogun-buildbot | build #415 of FC19 - libshogun is complete: Failure [failed compile test] Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/415 blamelist: Saurabh7 <saurabh.mahindre@gmail.com> | 18:50 |
shogun-buildbot | build #416 of FC19 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20libshogun/builds/416 | 18:54 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 260 seconds] | 19:03 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 19:03 | |
-!- travis-ci [~travis-ci@ec2-184-73-147-127.compute-1.amazonaws.com] has joined #shogun | 19:21 | |
travis-ci | [travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/16108828 | 19:21 |
-!- travis-ci [~travis-ci@ec2-184-73-147-127.compute-1.amazonaws.com] has left #shogun [] | 19:21 | |
-!- wiking [~wiking@info2k1.hu] has quit [Changing host] | 19:26 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 19:26 | |
-!- mode/#shogun [+o wiking] by ChanServ | 19:26 | |
shogun-buildbot | build #154 of clang34 - undefined behaviour analysis is complete: Failure [failed compile test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/154 blamelist: Saurabh7 <saurabh.mahindre@gmail.com> | 19:32 |
shogun-buildbot | build #155 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/155 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:43 |
shogun-buildbot | build #151 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/151 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Saurabh7 <saurabh.mahindre@gmail.com> | 19:59 |
-!- iglesiasg [~iglesiasg@211.Red-83-40-129.dynamicIP.rima-tde.net] has quit [Ping timeout: 272 seconds] | 20:30 | |
shogun-buildbot | build #167 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/167 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Saurabh7 <saurabh.mahindre@gmail.com> | 20:42 |
-!- HeikoS [~heiko@0541999e.skybroadband.com] has quit [Ping timeout: 272 seconds] | 20:58 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 21:35 | |
-!- HeikoS [~heiko@0541999e.skybroadband.com] has joined #shogun | 22:30 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 22:30 | |
-!- iglesiasg [~iglesiasg@211.Red-83-40-129.dynamicIP.rima-tde.net] has joined #shogun | 22:53 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 22:53 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 23:20 | |
shogun-notifier- | shogun: Fernando Iglesias :develop * daaaa0d / src/shogun/multiclass/KNN.cpp: https://github.com/shogun-toolbox/shogun/commit/daaaa0d2a8c993bc34eea8d5115dce459fe3e1e9 | 23:20 |
shogun-notifier- | shogun: Fix leak in KNN::store_model_features, duplicate increases ref count | 23:20 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 75cf24e / src/shogun/multiclass/KNN.cpp: https://github.com/shogun-toolbox/shogun/commit/75cf24e7aa8dfacff576c4bc31565b77afe4fd62 | 23:20 |
shogun-notifier- | shogun: Merge pull request #1800 from iglesias/develop | 23:20 |
shogun-notifier- | shogun: | 23:20 |
shogun-notifier- | shogun: Fix leak in KNN::store_model_features, duplicate increases ref count | 23:20 |
shogun-buildbot | build #2558 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2558 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 23:24 |
@HeikoS | iglesiasg: hi! | 23:25 |
@HeikoS | did this patch solve the problem? or was it an old one? | 23:25 |
@HeikoS | iglesiasg: anyway, cool! | 23:25 |
shogun-buildbot | build #2559 of deb1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2559 | 23:26 |
@iglesiasg | HeikoS, hey! | 23:27 |
@iglesiasg | HeikoS, nothing new there unfortunately | 23:27 |
@HeikoS | iglesiasg: ah ok | 23:27 |
@iglesiasg | it was the one corresponding to the last comments I wrote a couple of days ago | 23:27 |
@HeikoS | iglesiasg: thanks anyway for going over this | 23:27 |
@iglesiasg | it is cool | 23:27 |
@HeikoS | iglesiasg: I am not so sure about duplicate anymore anyways | 23:27 |
@HeikoS | what do you think about just ditching it? | 23:28 |
@HeikoS | and using clone instead | 23:28 |
@HeikoS | which clones the full thing | 23:28 |
@HeikoS | would that be a problem? | 23:28 |
@HeikoS | like infinite stacks or so? | 23:29 |
@HeikoS | subsets is also a question | 23:29 |
@iglesiasg | HeikoS, so for what I have understood from store_model_features | 23:29 |
@HeikoS | ah no that would not be goo | 23:29 |
@HeikoS | d | 23:29 |
@iglesiasg | the features are not actually copied, but shared | 23:29 |
@HeikoS | nono | 23:29 |
@HeikoS | they are copied | 23:29 |
@iglesiasg | clone would copy them I believe | 23:29 |
@HeikoS | only those who are needed for the copy | 23:29 |
@iglesiasg | mmm | 23:29 |
@iglesiasg | I think they are not copied :) | 23:29 |
@HeikoS | like for SVM only the SVs are copied | 23:30 |
@iglesiasg | the feature matrix is not copied | 23:30 |
@HeikoS | for SVM they definitely are | 23:30 |
@HeikoS | for others, I dont know | 23:30 |
@HeikoS | argh | 23:30 |
@HeikoS | maybe I should have a look before talking about things ;) | 23:30 |
@iglesiasg | I have looked into KNN for this, which is related to the example | 23:30 |
@HeikoS | ah | 23:30 |
@iglesiasg | haha no problem | 23:30 |
@iglesiasg | in any case, for me the problem is now something interesting | 23:31 |
@HeikoS | why? | 23:32 |
@iglesiasg | I don't understand how it is possibly leaking | 23:32 |
@HeikoS | ah ok | 23:32 |
@iglesiasg | since the refcount object is shared | 23:32 |
@HeikoS | maybe a good opportunity to review the store_model_features idea | 23:32 |
@iglesiasg | when the original features are freed, the other ones should be freed as well, I believe | 23:32 |
@iglesiasg | mmm hold on | 23:32 |
@iglesiasg | maybe not | 23:33 |
@iglesiasg | the pointer passed to unref is the one that might be freed | 23:33 |
@iglesiasg | but not any other pointer | 23:33 |
@iglesiasg | even if this other pointer may contain internally the same refcount object | 23:33 |
@iglesiasg | I guess that would explain it | 23:33 |
@HeikoS | yeah | 23:34 |
@HeikoS | when its NULL, no unref | 23:34 |
@iglesiasg | HeikoS, in any case, what I found rather surprising was that using this duplicate method, the duplicated object gets the same refcount object as the original | 23:34 |
@HeikoS | yeah that is not good | 23:34 |
@iglesiasg | I don't know if this is done on purpose | 23:34 |
@sonney2k | no it is a bug(tm) | 23:35 |
@iglesiasg | or it is just a consequence of using the copy constructor without really looking into that | 23:35 |
@iglesiasg | I see | 23:35 |
@iglesiasg | then well, an idea might be to take a look at the copy constructor of SGRefObject | 23:35 |
@HeikoS | iglesiasg: should not be like that | 23:35 |
@iglesiasg | since it may not make sense at all that different Shogun objects share the refcount | 23:36 |
@sonney2k | HeikoS, actually does duplicate() do a deep copy or a shallow one? | 23:36 |
@HeikoS | this method should create a plain object that contains the same data as the current view on the features object | 23:36 |
@HeikoS | that is what we need for store_model_features | 23:36 |
@HeikoS | sonney2k: shallow | 23:36 |
@iglesiasg | sonney2k, definitely not deep | 23:36 |
@HeikoS | it should just take the current view and copy it | 23:36 |
@HeikoS | sonney2k: but there are two modes: the default is that *all* features are copied | 23:37 |
@HeikoS | and the quick one is that only relevant ones are copied | 23:37 |
@HeikoS | sonney2k: we had a discussion about memory limitations a while ago (a few years) remember? | 23:37 |
@sonney2k | in this case it should inherit refcounts except for the new object right? | 23:37 |
@sonney2k | HeikoS, yeah but I forgot the exact details | 23:38 |
@HeikoS | sonney2k: it should just return a new object with its own ref-count | 23:38 |
@HeikoS | like a flat version of the underlying feature object | 23:38 |
@HeikoS | without any subsets etc | 23:38 |
@HeikoS | I think this is also different from copy constructor | 23:39 |
@sonney2k | HeikoS, no copy constructor should call duplicate then | 23:39 |
@HeikoS | sonney2k: mmh copy constructor should in fact also copy subsets etc | 23:40 |
@HeikoS | so thats different | 23:40 |
@HeikoS | duplicate should be called flat_copy() or so | 23:40 |
@HeikoS | btw | 23:41 |
@HeikoS | store_model_features should just save the current view | 23:42 |
@HeikoS | with views, no copying is needed | 23:42 |
@HeikoS | For x-validation that is really a design flaw ,we should have used multiple views from the beginning | 23:42 |
@iglesiasg | that agrees with what the operation of store_model_features in KNN | 23:42 |
@sonney2k | HeikoS, I think subsets should be copied to even in flat copy | 23:42 |
@HeikoS | because if the model contains some view, then the feature object itself can be shared | 23:43 |
@iglesiasg | at least how I understood it | 23:43 |
@HeikoS | among folds | 23:43 |
@HeikoS | sonney2k: yeah, I agree. But that is not what is needed for store_model_ | 23:43 |
@sonney2k | HeikoS, I mean back then we said shallow copy should keep all references of sub objects (and increase them) | 23:43 |
@sonney2k | so basically an object that is cheap to create | 23:43 |
@HeikoS | iglesiasg, sonney2k I think it might be a nice task for this guy who wants to help us to implement multiple subset/views and then use that in store_model_features | 23:44 |
@HeikoS | sonney2k: yeah | 23:44 |
@HeikoS | agree | 23:44 |
@sonney2k | but modifying data in it will modify it in all cheap clones | 23:44 |
@HeikoS | training usually doesnt modify right? | 23:44 |
@sonney2k | why not just do a drop subsets or so? | 23:44 |
@HeikoS | or does it? | 23:44 |
@HeikoS | sonney2k: ah yeah that is done in many methods | 23:44 |
@sonney2k | it can but usually not (preprocessers might...) | 23:44 |
@HeikoS | but think about this store_model_features | 23:44 |
@HeikoS | if it just stores a current view on a fixed fdeature object | 23:45 |
@HeikoS | everything is fine | 23:45 |
@HeikoS | since the reference is kept | 23:45 |
@HeikoS | so even if one assigns a different one for testing | 23:45 |
@HeikoS | the old model view on the old features object is still there | 23:45 |
@HeikoS | maybe worth thinking about those views a bit more | 23:45 |
@HeikoS | also should be unified for labels and features | 23:45 |
@sonney2k | 'real' views have the issue that one needs to modify all algorithms to properly work with them | 23:48 |
@sonney2k | as in get_feature_vector(i, view) I mean | 23:48 |
@HeikoS | sonney2k: no that would be annoying | 23:48 |
@HeikoS | all existing calls should use std view | 23:48 |
@HeikoS | and there should be a method to set the std view to a different one, this way whole algorithms can be changed | 23:49 |
@HeikoS | and all the x-validation stuff, and also wiking 's bagging can use the new ones | 23:49 |
@HeikoS | actually thats even next on my list on things to implement to the core | 23:49 |
@sonney2k | it is the most clean thing to do tough - thread safe and stacking and everythign would work with this right? | 23:49 |
@sonney2k | I know it is annoying that is why we are doing it the other way currenlty... | 23:50 |
@HeikoS | yeah | 23:51 |
@HeikoS | well, it is not too hard to change things in my eyes | 23:51 |
@HeikoS | and its somewhat nice to do | 23:51 |
@HeikoS | not as ugly as other improvements ;) | 23:52 |
@HeikoS | iglesiasg, we should push this indian guy to do it ;) | 23:52 |
@iglesiasg | taking advantage of man power yeah! | 23:53 |
@HeikoS | iglesiasg: hehe ;) | 23:54 |
@HeikoS | ok so we should brief him | 23:54 |
@HeikoS | what would be first step? | 23:55 |
@HeikoS | how should the interface to this look like? | 23:55 |
@iglesiasg | to store_model_features? | 23:56 |
@HeikoS | no to those multiple views/subsets | 23:57 |
@HeikoS | currently there is just one | 23:57 |
@HeikoS | and one can stack them | 23:57 |
@HeikoS | what we needed was a map of those stacks | 23:58 |
@HeikoS | with the key being some string or so | 23:58 |
@HeikoS | and then one would have to set a default subset stack | 23:58 |
@HeikoS | but every method to access features should be overloaded with an additional parameter that specified the view | 23:58 |
@HeikoS | or, alternatively, keep all calls the same and one has to change the current one by hand everytime (inconvinient) | 23:59 |
--- Log closed Mon Dec 30 00:00:29 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!