--- Log opened Mon Oct 19 00:00:09 2015 | ||
-!- PirosB3 [~pirosb3@172.56.20.1] has joined #shogun | 00:46 | |
-!- PirosB3 [~pirosb3@172.56.20.1] has quit [Ping timeout: 260 seconds] | 03:03 | |
-!- PirosB3 [~pirosb3@c-107-3-151-95.hsd1.ca.comcast.net] has joined #shogun | 03:12 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Read error: Connection reset by peer] | 03:15 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun | 03:15 | |
-!- PirosB3 [~pirosb3@c-107-3-151-95.hsd1.ca.comcast.net] has quit [Quit: PirosB3] | 04:33 | |
-!- PirosB3 [~pirosb3@c-24-4-56-73.hsd1.ca.comcast.net] has joined #shogun | 05:53 | |
-!- PirosB3 [~pirosb3@c-24-4-56-73.hsd1.ca.comcast.net] has quit [Quit: PirosB3] | 06:34 | |
-!- PirosB3 [~pirosb3@c-24-4-56-73.hsd1.ca.comcast.net] has joined #shogun | 06:35 | |
@wiking | lisitsyn: hey here? | 07:14 |
---|---|---|
-!- PirosB3 [~pirosb3@c-24-4-56-73.hsd1.ca.comcast.net] has quit [Quit: PirosB3] | 11:07 | |
-!- sonne|work [~sonnenbu@ip5b415e18.dynamic.kabel-deutschland.de] has joined #shogun | 13:30 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Read error: Connection reset by peer] | 14:52 | |
-!- shaochuan [~shaochuan@2601:647:4603:2a70:35a4:52ce:a5bd:a690] has joined #shogun | 14:53 | |
-!- shaochuan [~shaochuan@2601:647:4603:2a70:35a4:52ce:a5bd:a690] has quit [Remote host closed the connection] | 18:52 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun | 18:52 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Ping timeout: 250 seconds] | 18:57 | |
-!- sonney2k_ is now known as sonney2k | 20:48 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 20:48 | |
@sonney2k | lisitsyn, was wiking ever awake when you were? | 20:54 |
lisitsyn | sonney2k: yeah we catched up a few times | 20:55 |
@sonney2k | lisitsyn, around when? | 20:55 |
lisitsyn | sonney2k: closer to 'our' morning actually | 20:55 |
* sonney2k needs chocolate | 20:55 | |
lisitsyn | heh | 20:55 |
@sonney2k | nothing in the whole #&%*YfhDF house | 20:55 |
@sonney2k | I guess I could eat 1kg of it | 20:56 |
lisitsyn | hehe | 20:56 |
@sonney2k | jsut now | 20:56 |
lisitsyn | sonney2k: is he in singapore now? I have no idea to be honest | 20:56 |
lisitsyn | I guess so.. | 20:56 |
@sonney2k | yeah | 20:56 |
lisitsyn | I remember we chatting around 1-3 pm CET | 20:57 |
lisitsyn | so I guess if you want to catch up you gotta try tomorrow heh | 20:57 |
@sonney2k | lisitsyn, will try | 20:58 |
@sonney2k | but I don't actually recall how to write c++ code ;) | 20:58 |
lisitsyn | sonney2k: is it about license switch? | 20:58 |
@sonney2k | yeah | 20:59 |
lisitsyn | sonney2k: ask me I do that for living haha | 20:59 |
lisitsyn | why C++? | 20:59 |
@sonney2k | he wanted it badly but IMHO didn't really spent time pushing it | 20:59 |
lisitsyn | want to write some code? | 20:59 |
@sonney2k | well C/C++ all the same to me | 20:59 |
@sonney2k | shogun wise you mean | 20:59 |
@sonney2k | nah I am so java polluted now | 21:00 |
lisitsyn | yeah I mean why bother about C++ | 21:00 |
@sonney2k | I leak memory and write code w/o thinking | 21:00 |
@sonney2k | well shogun!~ | 21:00 |
@sonney2k | or what do you program | 21:00 |
@sonney2k | and which $LANG? | 21:00 |
lisitsyn | well you know shogun is really slow | 21:01 |
lisitsyn | if you're asking about me I am doing C++ and python | 21:01 |
@sonney2k | lisitsyn, shogun & slow you mean development currently? | 21:02 |
@sonney2k | or what? | 21:02 |
lisitsyn | sonney2k: yes basically no real development | 21:03 |
lisitsyn | well it is actually good if we were satisfied | 21:03 |
@sonney2k | lisitsyn, I am cleaning my fish tank ATM so I am not quick | 21:03 |
lisitsyn | but everyone has ideas and we don't move | 21:03 |
@sonney2k | lisitsyn, heh well I thought heiko gave up? | 21:03 |
lisitsyn | they sum to zero | 21:03 |
@sonney2k | well I guess no one has time right? | 21:04 |
@sonney2k | or everyone wants to do something different :) | 21:04 |
@sonney2k | lisitsyn, so you use scikits now or what? | 21:05 |
@sonney2k | lisitsyn, and what would you improve if you had the time? | 21:05 |
lisitsyn | sonney2k: sorry, back | 21:06 |
lisitsyn | sonney2k: I wouldn't say its about time | 21:06 |
lisitsyn | some kind of passion I'd say | 21:06 |
@sonney2k | yeah if you don't use it why do it... | 21:06 |
@sonney2k | lisitsyn, what do you use now? | 21:06 |
lisitsyn | I am doing some neural nets now (fuuu) | 21:06 |
lisitsyn | with in-house stuff | 21:06 |
lisitsyn | big company has their own tools | 21:07 |
lisitsyn | but I used shogun a few times | 21:07 |
lisitsyn | we need to improve how we handle data and modularize | 21:07 |
@sonney2k | lisitsyn, handle data as in? | 21:07 |
lisitsyn | well lacks some magic | 21:07 |
lisitsyn | have you used vw? | 21:08 |
@sonney2k | lisitsyn, never | 21:08 |
lisitsyn | it has strict format but you can do basically anything | 21:08 |
lisitsyn | there is no such thing as dense features or sparse features in real life | 21:08 |
lisitsyn | they are always mixed together | 21:08 |
@sonney2k | with the size of data I had to deal with I always had to write the basic i/o from scratch... | 21:08 |
@sonney2k | well I manually de-mixed the features back in the days | 21:09 |
lisitsyn | yeah | 21:09 |
@sonney2k | but regarding modularize | 21:09 |
lisitsyn | but things speed up | 21:09 |
lisitsyn | and you can't spend your time doing that | 21:09 |
@sonney2k | I see that this could in principle be done for libshogun / c++ | 21:09 |
@sonney2k | but I don't see it for *any* interface | 21:09 |
@sonney2k | at least not with swig | 21:10 |
lisitsyn | have you seen my mail? | 21:10 |
lisitsyn | I had some solution | 21:10 |
@sonney2k | nah I don't read email | 21:10 |
lisitsyn | pretty ugly.. | 21:10 |
lisitsyn | quite similar to static | 21:10 |
@sonney2k | I am only scanning things... | 21:10 |
lisitsyn | well if you have parameter just set it with set("parameter", 3); | 21:10 |
lisitsyn | like that | 21:10 |
lisitsyn | or other thing, tags | 21:11 |
lisitsyn | like you have tags for parameters | 21:11 |
@sonney2k | yeah sounds like static | 21:11 |
lisitsyn | tag = kernel.parameter("width"); | 21:11 |
lisitsyn | kernel.set(tag, 3.0); | 21:11 |
lisitsyn | this would work in swig | 21:11 |
@sonney2k | ahh you are talking about this still being wrapped in swig | 21:12 |
lisitsyn | yes | 21:12 |
@sonney2k | I see | 21:12 |
lisitsyn | w/o swig doesn't make sense | 21:12 |
@sonney2k | but still you would need to expose the whole object .h tree to swig | 21:12 |
@sonney2k | and you cannot modularize that | 21:12 |
lisitsyn | no, you can avoid that | 21:12 |
@sonney2k | how? | 21:12 |
lisitsyn | you can expose only basic objects | 21:12 |
lisitsyn | classifier, kernel, etc | 21:12 |
lisitsyn | and call kernel("gaussian") | 21:12 |
@sonney2k | you would need to feed any object into another | 21:12 |
lisitsyn | well that's the limitation, you would have to use only base classes | 21:13 |
lisitsyn | no crazy relationship between two really different modules | 21:13 |
@sonney2k | could you please make an example for kernel and features | 21:13 |
@sonney2k | so I know how it could work? | 21:14 |
lisitsyn | hmm | 21:14 |
lisitsyn | well, features are base classes | 21:14 |
lisitsyn | features = shogun.DenseFeatures would stay I guess | 21:14 |
lisitsyn | as for kernel you just do | 21:14 |
lisitsyn | kernel = shogun.kernel("gaussian") | 21:14 |
lisitsyn | kernel.init(features, features) | 21:15 |
lisitsyn | kernel.set("width", 10.0) | 21:15 |
lisitsyn | sonney2k: most checks would go to runtime then | 21:15 |
lisitsyn | but I don't see any need to expose .h of plugins this way | 21:15 |
lisitsyn | classifier = shogun.classifier("liblinear") | 21:16 |
lisitsyn | classifier.train(kernel) | 21:16 |
@sonney2k | well ok features make up 50% of all swig wrapped code | 21:18 |
lisitsyn | sonney2k: I would actually drop features.. | 21:19 |
lisitsyn | and write adapters to use pandas dataframe | 21:19 |
lisitsyn | spark rdd | 21:19 |
lisitsyn | and stuff like that | 21:19 |
@sonney2k | lisitsyn, so you would have a classifier | 21:20 |
@sonney2k | sure you can init is just by some 'ascii' name | 21:20 |
lisitsyn | yes | 21:20 |
@sonney2k | so you would just have a stand alone classifier module | 21:20 |
lisitsyn | yeah it would be easier to write/load plugins then | 21:21 |
lisitsyn | sonney2k: I am also obsessed by other thing | 21:21 |
@sonney2k | but how do classifiers underneath work with features w/o knowing about e.g. CDenseFeatures? | 21:21 |
lisitsyn | why? | 21:21 |
lisitsyn | it knows about dense features | 21:21 |
lisitsyn | they would have to be in the base shogun | 21:21 |
lisitsyn | it doesn't know about gaussian kernel | 21:21 |
@sonney2k | lisitsyn, OK | 21:21 |
lisitsyn | or oligowdstringkernelwhatever | 21:21 |
lisitsyn | sonney2k: I am obsessed about having shogun.classifier('some_matlab_classifier_from_fancy_researcher') | 21:22 |
@sonney2k | but then kernelmachines would need to know about kernels right? | 21:22 |
lisitsyn | yes about base class, CKernel | 21:22 |
@sonney2k | ok makes sense | 21:22 |
@sonney2k | so basically you modularize out the machine part right? | 21:22 |
lisitsyn | we would just move all the concrete kernels | 21:22 |
lisitsyn | concrete machines | 21:23 |
lisitsyn | and stuff like that | 21:23 |
lisitsyn | this would be a step | 21:23 |
@sonney2k | this sounds very doable | 21:23 |
lisitsyn | yeah probably | 21:23 |
lisitsyn | but this string-based thing makes me cry and I see no other way | 21:23 |
lisitsyn | when it comes to something dynamic all you have is some string | 21:23 |
@sonney2k | lisitsyn, yeah even enums would make it non-dynamic | 21:24 |
lisitsyn | sonney2k: and we should employ your COFFIN much more | 21:26 |
lisitsyn | sonney2k: recently I realized nobody use that approach for neural nets | 21:26 |
lisitsyn | although it is quite natural | 21:26 |
@sonney2k | lisitsyn, so if I get it right you would have a dynamic runtime open for the classifier | 21:26 |
lisitsyn | yes | 21:26 |
@sonney2k | and this way you can just have ton's of .so's hanging around | 21:27 |
@sonney2k | and nice and modular and you only need the libshogun core installed | 21:27 |
@sonney2k | lisitsyn, and if you want to use more of shogun's helper stuff linalg etc? | 21:28 |
lisitsyn | not sure | 21:28 |
lisitsyn | sonney2k: I am not sure anybody needs it being exposed via swig | 21:28 |
lisitsyn | all major langs have their own linalg | 21:28 |
@sonney2k | lisitsyn, well not swig but from the c++ side | 21:29 |
@sonney2k | you would still need to link against that part | 21:29 |
lisitsyn | ah | 21:30 |
lisitsyn | true | 21:30 |
@sonney2k | lisitsyn, and yes about COFFIN... if I use stuff I only do it linearly and the coffin way | 21:30 |
lisitsyn | sonney2k: it is a very good idea, essentially employing principle of limiting the interface | 21:31 |
lisitsyn | like you don't need all the stuff so don't ask for that broad interface | 21:31 |
@sonney2k | lisitsyn, thanks ;) | 21:34 |
@sonney2k | lisitsyn, the one thing that feels hard with c++ to me is I have to use my brain to not leak | 21:35 |
@sonney2k | lisitsyn, I really cannot do that (by default) | 21:35 |
lisitsyn | sonney2k: ah yes that's another thing to change | 21:36 |
lisitsyn | I'd like to not see any raw pointer in shogun | 21:36 |
lisitsyn | sonney2k: you know it is quite natural for modern C++ to use unique and shared pointer semantics | 21:38 |
lisitsyn | *sometimes* weak | 21:38 |
lisitsyn | but that's all to not think a lot | 21:38 |
@sonney2k | lisitsyn, alright and how do we get it in then? | 22:11 |
@sonney2k | lisitsyn, this is really a blocker to me. the modularization part sounds rather easy to do and could even be done in a compatible way in *addition* to what we have now until it is funcitonal | 22:12 |
@wiking | ho | 22:13 |
@sonney2k | wiking, hey - lets chew bubble gum and kick ass... | 22:16 |
@sonney2k | wiking, well I guess I will have to chew it for you :P | 22:16 |
@sonney2k | wiking, so where the heck is the license thing?! | 22:16 |
lisitsyn | sonney2k: that's true | 22:23 |
lisitsyn | we can start right away heh | 22:23 |
lisitsyn | no idea what blocks me | 22:23 |
@wiking | :>>>> | 22:25 |
@wiking | with one of the website deploys | 22:25 |
@wiking | that thing got wiped out | 22:25 |
@wiking | and as much as i wanted couldnt find any copy of it anywhere | 22:25 |
lisitsyn | lol | 22:25 |
@wiking | and i have no idea what i put on that page :D | 22:25 |
@wiking | i mean text wise | 22:25 |
@sonney2k | wiking, and I guess no idea who signed it either? | 22:26 |
@sonney2k | wiking, man! | 22:26 |
@sonney2k | wiking, doesn't sound like you pushed very hard didn't you? | 22:26 |
@wiking | nono | 22:26 |
@wiking | that i know | 22:26 |
@wiking | :) | 22:26 |
lisitsyn | lets do that once again lool | 22:26 |
@wiking | hahaha you think of me too little ;) | 22:26 |
@wiking | http://maeth.com/shogun_licensing_stats.php | 22:26 |
@wiking | that's in a db table | 22:27 |
lisitsyn | ah | 22:27 |
@wiking | i just need the stupid static page | 22:27 |
lisitsyn | so you have a list of signers | 22:27 |
lisitsyn | but no text | 22:27 |
@wiking | with one webform in it | 22:27 |
lisitsyn | that's funny | 22:27 |
@wiking | i just need to recover the webform basically | 22:27 |
@wiking | that'll send some POST params to a php script | 22:28 |
@wiking | that'll update this table | 22:28 |
@wiking | so stop being cocky and help me with the content of the webform :) | 22:28 |
@sonney2k | wiking, didn't you send out an emal to everyone? | 22:31 |
@sonney2k | I mean that content should be in everyones inbox | 22:32 |
@wiking | yeah partly in some way yes | 22:32 |
@sonney2k | wiking, do you remember the sender | 22:34 |
lisitsyn | Dear root, | 22:35 |
lisitsyn | We are contacting you as You are one of the developers who has been committing to Shogun machine learning toolbox in the past. | 22:35 |
lisitsyn | like that? | 22:36 |
lisitsyn | wiking: we your inbox | 22:37 |
lisitsyn | see | 22:37 |
@sonney2k | lisitsyn, subject? | 22:44 |
lisitsyn | Shogun Licensing | 22:44 |
@sonney2k | lisitsyn, found ... | 22:45 |
@sonney2k | wiking, was the webform on shogun-toolbox.org's servers? | 22:46 |
@sonney2k | because the url I got was | 22:46 |
@sonney2k | wiking, http://shogun-toolbox.org/static/license/ | 22:47 |
@wiking | yes | 22:48 |
@sonney2k | wiking, well we have (or had) backups | 22:50 |
@sonney2k | wiking, but probably not dating back 2 years | 22:50 |
* sonney2k Zzzzz | 22:52 | |
--- Log closed Tue Oct 20 00:00:10 2015 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!