--- Log opened Wed Mar 12 00:00:35 2014 | ||
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 00:26 | |
-!- hushell [~hushell@8-12.ptpg.oregonstate.edu] has joined #shogun | 01:06 | |
hushell | hi guys, ask a question about IO | 01:07 |
---|---|---|
hushell | can we use fstream in arguments of some member functions? | 01:07 |
thoralf | hushell: Hey. | 01:13 |
thoralf | hushell: AFAIK they don't like STL at all. | 01:14 |
hushell | thoralf: :) | 01:28 |
hushell | thoralf: so how about use std:ifstream inside? ;) I don't want to use CXXXFile, we have CLineReader, but there is no LineWriter | 01:29 |
shogun-buildbot_ | build #1596 of rpm1 - libshogun is complete: Failure [failed compile test] Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1596 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>, frank0523 <zhangyuyu2008@gmail.com> | 01:30 |
thoralf | hushell: I know. | 01:30 |
hushell | thoralf: in the arguments, only ask for a char* to specify the path | 01:30 |
thoralf | hushell: If you use it inside the implementation, it should be fine. | 01:30 |
thoralf | i.e. cpp file | 01:30 |
hushell | thoralf: I have experience with STL inside | 01:31 |
hushell | thoralf: thanks! I'll update my entrance task in that way | 01:31 |
thoralf | Yeah, writing line-based ascii data is just too simple to wrap it ten times. | 01:35 |
thoralf | https://github.com/shogun-toolbox/shogun/pull/1958/files ->src/shogun/labels/MultilabelLabels.cpp, line 358, method save() | 01:35 |
thoralf | ;) | 01:35 |
hushell | thoralf: exactly, simple and effective, I got confused by so many io classes | 01:42 |
-!- lisitsyn1 [~lisitsyn@80.252.20.67] has joined #shogun | 01:58 | |
-!- lisitsyn [~lisitsyn@80.252.20.67] has quit [Ping timeout: 240 seconds] | 01:59 | |
-!- gif [~gif@130.195.253.30] has quit [Ping timeout: 240 seconds] | 01:59 | |
-!- gif [~gif@130.195.253.30] has joined #shogun | 01:59 | |
-!- naywhayare [~ryan@spoon.lugatgt.org] has quit [Ping timeout: 240 seconds] | 02:05 | |
-!- naywhayare [~ryan@spoon.lugatgt.org] has joined #shogun | 02:05 | |
-!- gif [~gif@130.195.253.30] has quit [Remote host closed the connection] | 02:10 | |
abinash_panda | hushell: Hi | 02:34 |
thoralf | Hey abinash_panda | 02:35 |
abinash_panda | thoralf: Hi | 02:35 |
abinash_panda | thoralf: So, I got the mail of hushell | 02:36 |
abinash_panda | thoralf: Should I be sending the same PR again | 02:36 |
shogun-buildbot_ | build #217 of debian wheezy - memcheck is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/217 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>, frank0523 <zhangyuyu2008@gmail.com> | 02:36 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:36 | |
thoralf | abinash_panda: No, I think you should change your PR a bit. You should either use CLineReader or CCSVReader in your load() method. | 02:37 |
abinash_panda | thoralf: Ok. | 02:38 |
thoralf | It's 2:37 am. | 02:38 |
thoralf | Need to sleep. | 02:38 |
abinash_panda | thoralf: Could you please specify where should I use them. | 02:38 |
abinash_panda | thoralf: Ok. Fine. I would find it on my own. :) | 02:39 |
thoralf | Hmm. | 02:39 |
thoralf | abinash_panda: You didn't implement a load() method yet. I see. | 02:39 |
abinash_panda | thoralf: Yeah. I have only implemented the save method. | 02:40 |
abinash_panda | thoralf: Should I go for the implementation of load() method or should I wait for the comments regarding the save() method | 02:41 |
thoralf | abinash_panda: You can do that - but be prepared to move the method to whereever it should stay finally. | 02:42 |
abinash_panda | thoralf: Yeah. Fine. | 02:42 |
abinash_panda | thoralf: Sending PR right now. | 02:42 |
thoralf | I think technically it won't change much, but it's still not clear where to put it. | 02:42 |
abinash_panda | thoralf: Ok. Thanks for helping me out :) | 02:43 |
thoralf | Maybe you can think about how to extract load()/save() and where to put it and make a suggestion on the list? | 02:43 |
abinash_panda | thoralf: Ok. I would be making suggestions on the list regarding the methods. | 02:45 |
abinash_panda | thoralf: Thanks again and good night :) | 02:47 |
thoralf | Thanks :) | 02:47 |
-!- thoralf [~thoralf@91-65-142-206-dynip.superkabel.de] has quit [Quit: Konversation terminated!] | 02:48 | |
-!- hushell [~hushell@8-12.ptpg.oregonstate.edu] has quit [Ping timeout: 245 seconds] | 02:57 | |
-!- gif [~gif@130.195.253.30] has joined #shogun | 03:04 | |
-!- gif [~gif@130.195.253.30] has quit [Client Quit] | 03:08 | |
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 264 seconds] | 03:21 | |
-!- abinash_panda [01261510@gateway/web/freenode/ip.1.38.21.16] has quit [Quit: Page closed] | 03:31 | |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has joined #shogun | 03:41 | |
-!- zxtx [~zv@129-79-241-148.dhcp-bl.indiana.edu] has quit [Ping timeout: 246 seconds] | 04:07 | |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has quit [Ping timeout: 246 seconds] | 04:14 | |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has joined #shogun | 04:21 | |
-!- zxtx [~zv@c-98-223-196-32.hsd1.in.comcast.net] has joined #shogun | 05:06 | |
-!- kprah [67157e4c@gateway/web/freenode/ip.103.21.126.76] has joined #shogun | 05:12 | |
-!- kprah [67157e4c@gateway/web/freenode/ip.103.21.126.76] has quit [Client Quit] | 05:12 | |
-!- abinash_panda [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has joined #shogun | 05:53 | |
abinash_panda | hushell: Hi ! | 05:54 |
hushell | hi abinash_panda | 05:59 |
hushell | abinash_panda: sorry I was busy in daytime | 06:00 |
abinash_panda | hushell: Ok. So can we discuss now? | 06:00 |
hushell | abinash_panda: sure thing | 06:00 |
hushell | abinash_panda: as I said in email, the stored file should be different from UAI format | 06:00 |
abinash_panda | Ok :). I want to clear few things. Should I go for std::ifstream or FILE* (I am confused) | 06:01 |
hushell | I'll make the format more clear in the task description | 06:01 |
hushell | abinash_panda: yeah, std::ifstream is good | 06:01 |
abinash_panda | as I have seen in most of the implementation of shogun, there is no use of std::ifstream or any STL | 06:01 |
abinash_panda | hushell: Ok. Fine | 06:02 |
hushell | look at L420 at https://github.com/shogun-toolbox/shogun/pull/1958/files | 06:02 |
hushell | this is Thoralf's implementation for his loading/saving | 06:02 |
abinash_panda | so if I am going to use std::ifstream, do I need to use CLineReader | 06:02 |
hushell | abinash_panda: well, inside a function is ok | 06:02 |
abinash_panda | Yeah. I have seen the implementation of Thoralf's | 06:03 |
hushell | abinash_panda: I don't suggest you to use LineReader | 06:03 |
hushell | abinash_panda: as you may notice, there is no linewriter | 06:03 |
abinash_panda | hushell: Ok. Fine. Its a relief. | 06:03 |
abinash_panda | I was very confused on how to use LineReader :) | 06:04 |
hushell | abinash_panda: I used STL in BeliefPropagation a lot | 06:04 |
abinash_panda | hushell: I will have a look at it. | 06:04 |
hushell | abinash_panda: so you want to keep working on the loading/saving? | 06:04 |
abinash_panda | hushell: Yeah. I have done loading method also. | 06:04 |
abinash_panda | but that uses FILE* | 06:05 |
abinash_panda | so I have to refactor that one, and I would send a PR | 06:05 |
hushell | abinash_panda: no rush, let me specify the format of our file | 06:06 |
abinash_panda | hushell: Yeah sure | 06:06 |
hushell | maybe we can have another function to convert to UAI format | 06:06 |
abinash_panda | Sorry, I couldn't get you | 06:07 |
hushell | abinash_panda: so could you say something about your background on PGM? | 06:07 |
hushell | abinash_panda: I mean the stored file will not be a in UAI format, as we will have factor type | 06:07 |
abinash_panda | hushell: I have been studying PGM on coursera. I also author of a PGM library in python called pgmpy. | 06:08 |
abinash_panda | https://github.com/pgmpy/pgmpy | 06:08 |
hushell | but we should also consider UAI format | 06:08 |
hushell | how you do inference in pgmpy? | 06:08 |
abinash_panda | as the library is very young, currently we have implemented only variable elimination for inference. | 06:09 |
abinash_panda | We are planning to extend this one in this year GSoC (as we have been selected for GSoC under PSF) | 06:10 |
hushell | abinash_panda: cool | 06:12 |
abinash_panda | hushell: so you were discussing about the new format | 06:14 |
hushell | abinash_panda: could you point me your implementation of VE? | 06:14 |
hushell | abinash_panda: yeah, we will change a bit the specification of function tables | 06:14 |
abinash_panda | hushell: Ok. Initially we created functions for factor multiplication using cython. | 06:15 |
abinash_panda | Then we used sum product VE for computing conditional probabilities | 06:15 |
hushell | abinash_panda: you may consider to merge yours with https://github.com/pystruct/pystruct | 06:16 |
abinash_panda | Ok. I would have a look at it. | 06:17 |
hushell | abinash_panda: do you have example for testing VE? | 06:17 |
abinash_panda | Thanks for pointing it out :) | 06:17 |
hushell | I'd like to have a try | 06:17 |
hushell | abinash_panda: your lib is good for teaching purpose :) | 06:18 |
hushell | BTW, do you know anything about structured output SVM? | 06:20 |
abinash_panda | Ok. I don't think this VE algorithm has been completely implemented yet. I would ask someone or I would complete myself ASAP. | 06:21 |
abinash_panda | hushell: No. I am a beginner in PGM. If you could point out some sources for understanding it. | 06:21 |
hushell | abinash_panda: no worry. I was searching for it :) I think you'll make it | 06:21 |
abinash_panda | hushell: Yeah. I would complete it ASAP :) | 06:22 |
hushell | abinash_panda: let me think, I can send you some papers and video tutorials | 06:22 |
abinash_panda | hushell: Yeah sure. That would be great. | 06:23 |
abinash_panda | hushell: I was thinking of working on the idea Structured Output Learning with Approximate Inferences for GSoC | 06:25 |
hushell | abinash_panda: well, I started from this paper one and half year ago: http://www.cs.cornell.edu/People/tj/publications/joachims_etal_09a.pdf | 06:25 |
hushell | but if this is your first time read a paper, it would be hard, but try to go through it | 06:26 |
hushell | abinash_panda: http://videolectures.net/mlg08_joachims_sop/ | 06:26 |
abinash_panda | Ok. I would go through the lectures. | 06:27 |
hushell | this video is helpful as well | 06:27 |
abinash_panda | As I have been working on PGM, so I thought that this idea would be great to work on. | 06:27 |
abinash_panda | I would go through the lectures and read papers on them. I am fast learner :) | 06:28 |
hushell | abinash_panda: the approximate inference idea is based on what I did last year | 06:28 |
hushell | abinash_panda: honestly, it would be a bit too hard for someone new to SO learning, but I encourage you to have a try | 06:28 |
abinash_panda | hushell: Yeah. I would have a try. | 06:29 |
abinash_panda | let see upto what extent I am sucesfull :) | 06:29 |
hushell | abinash_panda: sounds great! looking forward your application! | 06:31 |
hushell | abinash_panda: BTW, keep working on the entrance tasks | 06:31 |
abinash_panda | cool. I would be drafting my application after I would go through the topics. | 06:31 |
hushell | I prepared another multi-label classification task, you may also see it | 06:31 |
abinash_panda | hushell: Please guide me through the application process :) | 06:31 |
hushell | abinash_panda: my suggestion is finish some tasks first, and write a good proposal to introduce yourself and your plan for this summer | 06:32 |
hushell | let me find my application last year | 06:33 |
abinash_panda | Cool | 06:33 |
abinash_panda | so I should finish some entrance task first, right ? | 06:33 |
hushell | abinash_panda: It's not necessary to finish some, but would be helpful | 06:35 |
-!- travis-ci [~travis-ci@ec2-54-205-8-230.compute-1.amazonaws.com] has joined #shogun | 06:35 | |
travis-ci | [travis-ci] it's Fernando Iglesias'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/20567508 | 06:35 |
-!- travis-ci [~travis-ci@ec2-54-205-8-230.compute-1.amazonaws.com] has left #shogun [] | 06:35 | |
abinash_panda | I got it. | 06:35 |
abinash_panda | can you provide me the link to the new multi-label classification task | 06:35 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 06:35 | |
shogun-notifier- | shogun: Parijat Mazumdar :develop * a551ff2 / src/shogun/multiclass/tree/ (2 files): https://github.com/shogun-toolbox/shogun/commit/a551ff2fca6f9ee37bf803dafce20543ff2105c6 | 06:35 |
shogun-notifier- | shogun: id3 tree added | 06:35 |
shogun-notifier- | shogun: Parijat Mazumdar :develop * 8712416 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/8712416e60c0a2e259c60f09c75dfe50f2219a73 | 06:35 |
shogun-notifier- | shogun: unittest added + brief added | 06:35 |
shogun-notifier- | shogun: Parijat Mazumdar :develop * 91e2b5c / src/shogun/multiclass/tree/ID3ClassifierTree.h: https://github.com/shogun-toolbox/shogun/commit/91e2b5c8c01974771a04c17cfeb492e8900e69fd | 06:35 |
shogun-notifier- | shogun: algo added in brief | 06:35 |
shogun-notifier- | shogun: Parijat Mazumdar :develop * 87ca78d / src/shogun/multiclass/tree/ID3ClassifierTree.cpp: https://github.com/shogun-toolbox/shogun/commit/87ca78d072e4bb71fed8913b3c77de4e94365c81 | 06:35 |
shogun-notifier- | shogun: entropy method updated | 06:35 |
shogun-notifier- | shogun: Parijat Mazumdar :develop * 41d3666 / / (2 files): https://github.com/shogun-toolbox/shogun/commit/41d3666449cc79c3549de632994e57f60ad3cf94 | 06:35 |
shogun-notifier- | shogun: added remaining case in unittest + code style changes | 06:35 |
shogun-notifier- | shogun: Parijat Mazumdar :develop * c34fd60 / NEWS: https://github.com/shogun-toolbox/shogun/commit/c34fd609a85d2a6909bf1328209833d916711cc5 | 06:35 |
shogun-notifier- | shogun: news updated | 06:35 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 52b595e / / (4 files): https://github.com/shogun-toolbox/shogun/commit/52b595e6a7b1108da544f71419effa79e10523cd | 06:35 |
shogun-notifier- | shogun: Merge pull request #1970 from mazumdarparijat/pca | 06:35 |
hushell | abinash_panda: sorry I couldn't find mine, you may search in shogun's website to find some templates | 06:37 |
abinash_panda | Ok. I would find it. | 06:37 |
hushell | abinash_panda: https://github.com/shogun-toolbox/shogun/issues/1916 | 06:37 |
hushell | here is the multi-label task | 06:37 |
abinash_panda | Yeah. I would look into it. | 06:38 |
hushell | abinash_panda: I have to work on my project now, talk you later and good luck! | 06:38 |
abinash_panda | By the mean time, you could decide the file format | 06:38 |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has quit [Quit: WeeChat 0.3.7] | 06:38 | |
abinash_panda | hushell: Bye :) | 06:38 |
-!- vasanth [~vasanth@203.129.242.82] has joined #shogun | 06:56 | |
shogun-buildbot_ | build #1597 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1597 | 07:07 |
-!- vasanth [~vasanth@203.129.242.82] has quit [Quit: Konversation terminated!] | 07:27 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 07:55 | |
wiking_ | fucking hell | 08:09 |
-!- wiking_ is now known as wiking | 08:09 | |
wiking | what is this disobeyance in shogun? | 08:10 |
sonne|work | wiking: what? | 08:10 |
wiking | sonne|work: about this UAIFile ... | 08:10 |
wiking | i really do not understand why cannot people grasp that we want file handling in IO | 08:11 |
wiking | and not in a random fucking class | 08:11 |
wiking | and now even hushell is encouring the random implementation... | 08:11 |
wiking | i dont get this.. | 08:11 |
lisitsyn1 | haha | 08:14 |
wiking | lisitsyn1: i bet you like my email :D | 08:14 |
lisitsyn1 | hahah | 08:15 |
lisitsyn1 | wiking: terrific | 08:15 |
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has quit [Ping timeout: 265 seconds] | 08:17 | |
wiking | lisitsyn1: a bit more and i'll go with the passive agressive 'my 2 cents' shit :) | 08:17 |
wiking | or should i make more fun of | 08:17 |
wiking | 'CSV, SVMLight, HDF5 is a very specific file format as well' | 08:18 |
wiking | still we have IO class for that... | 08:18 |
lisitsyn1 | wiking: it continues | 08:34 |
lisitsyn1 | ;) | 08:34 |
wiking | omfg | 08:36 |
wiking | done | 08:41 |
lisitsyn1 | haha amazing | 08:43 |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 08:53 | |
-!- fs_ [~floriansc@g226194186.adsl.alicedsl.de] has joined #shogun | 08:57 | |
fs_ | hi | 08:58 |
sonne|work | hi | 08:58 |
-!- wiking [~wiking@info2k1.hu] has quit [Changing host] | 09:02 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 09:02 | |
-!- mode/#shogun [+o wiking] by ChanServ | 09:02 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 09:35 | |
shogun-buildbot_ | build #218 of debian wheezy - memcheck is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/218 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>, Parijat Mazumdar <mazumdarparijat@gmail.com> | 10:13 |
@wiking | lisitsyn1: saga continues :D | 10:28 |
lisitsyn1 | wiking: wherE? | 10:29 |
@wiking | lisitsyn1: https://github.com/shogun-toolbox/shogun/issues/1913#issuecomment-37388247 | 10:30 |
lisitsyn1 | haha | 10:30 |
@wiking | seriously i do not understand what i'm writing wrong that people seems to still not understand what i'm trying to say | 10:31 |
@wiking | lisitsyn1: do you get what i want? :D or you dont understand it either | 10:31 |
lisitsyn1 | wiking: well yeah uai is a file and we have classes for various files | 10:31 |
lisitsyn1 | makes sense to have a class for this file as well :D | 10:32 |
@wiking | :> | 10:32 |
@wiking | i think what's missing here in the head the abstraction step that decouples a given representation (factorgraph) from a file format | 10:34 |
@wiking | and to see why it's better to have something in a structured format in memory with the right API before actually instanciating an object | 10:36 |
@wiking | i think it should be mandatory that everybody learns OSI :) | 10:36 |
@wiking | lisitsyn1: https://github.com/shogun-toolbox/shogun/issues/1913#issuecomment-37389931 | 10:39 |
lisitsyn1 | wiking: heh | 10:42 |
-!- PirosB3 [~pirosb3@162.219.6.191] has joined #shogun | 11:19 | |
PirosB3 | hey guys :) | 11:19 |
PirosB3 | I am having a strange error compiling on Ubuntu: http://pastebin.instantbird.com/630705 | 11:21 |
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has joined #shogun | 11:21 | |
PirosB3 | has anyone had this error before? I found some comments on the ML but I didn't find a solution to the problem | 11:21 |
sonne|work | PirosB3: yeah as it says - you need to compile with -fPIC | 11:22 |
PirosB3 | sonne|work: so I will just do make -j3 -fPIC ? | 11:22 |
sonne|work | PirosB3: how do you compile in the first place? -fPIC is automagically detected... | 11:23 |
PirosB3 | sonne|work: just as instructed on INSTALL. cd build && cmake params .. && make -j3 | 11:23 |
sonne|work | and which shogun version? | 11:24 |
PirosB3 | develop | 11:24 |
sonne|work | so current git? | 11:24 |
PirosB3 | yep | 11:25 |
sonne|work | PirosB3: then please do VERBOSE=1 make | 11:25 |
PirosB3 | okay | 11:25 |
sonne|work | and paste the output somewhere | 11:25 |
PirosB3 | CMakeFiles/libshogun.dir/lib/SGVector.cpp.o: could not read symbols: Bad value | 11:25 |
PirosB3 | what is this?! | 11:25 |
PirosB3 | ok pulled latest develop and recompiling everything | 11:27 |
PirosB3 | sonne|work: how much time does it take for you to compile?! I am going crazy. every time it's +30m! | 11:27 |
sonne|work | PirosB3: install ccache | 11:27 |
@wiking | PirosB3: do u have ccache installed? | 11:27 |
PirosB3 | nope | 11:27 |
PirosB3 | will do immediately! | 11:28 |
PirosB3 | do I need to link it up in any way? | 11:28 |
@wiking | no | 11:29 |
@wiking | just install it | 11:29 |
@wiking | cmake will pick it up | 11:29 |
PirosB3 | yep done :) | 11:33 |
PirosB3 | am I safe to run with -j3 ? | 11:33 |
PirosB3 | also, sonne|work should I do -fPIC? | 11:33 |
PirosB3 | make -fPIC | 11:33 |
@wiking | PirosB3: r u running this in a virtual machine? | 11:33 |
PirosB3 | wiking: yes | 11:34 |
PirosB3 | will not compile on my mac, too old OS :( | 11:34 |
@wiking | PirosB3: well i guess that's why u have problem with PIC | 11:34 |
PirosB3 | wiking: so do you suggest to compile with fPIC? | 11:34 |
@wiking | well as it says you'll have to | 11:35 |
@wiking | i guess | 11:35 |
@wiking | CFLAGS="-fPIC" CXXFLAGS="-fPIC" make | 11:35 |
@wiking | should do it | 11:35 |
PirosB3 | wiking: I did not perfectly understand what PIC is | 11:35 |
PirosB3 | also, it is okay to develop with PIC? or could it make my sistem different in any way? | 11:36 |
@wiking | PirosB3: http://lmgtfy.com/?q=fPIC | 11:36 |
PirosB3 | wiking: I already have | 11:36 |
@wiking | so then i guess you know what -fPIC is | 11:37 |
PirosB3 | wiking: I did not understand what position independent code is | 11:37 |
lisitsyn1 | that's about libraries and how you link them together PirosB3 | 11:37 |
@wiking | the first link on google has a pretty good description about what PIC is | 11:38 |
@wiking | PirosB3: what do you mean by "it is okay to develop with PIC?" ? as you can see you cannot even compile the library w/o PIC | 11:39 |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 11:41 | |
PirosB3 | wiking: I meant that if I develop with PIC, will it somehow make my environment different than the one that I would have compiling without the flag? | 11:41 |
lisitsyn1 | PirosB3: no | 11:41 |
PirosB3 | thx | 11:42 |
-!- jiaolong [9e6d09e4@gateway/web/freenode/ip.158.109.9.228] has joined #shogun | 11:52 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 11:58 | |
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 245 seconds] | 11:58 | |
fs_ | wiking, regarding decision trees: I'll take a look at everything. I'll get back to you in the evening (UTC). | 11:59 |
@wiking | fs_: that's alright... take your time. and please use mailing list or issue as a main communication since it has higher publicity ;) | 12:07 |
fs_ | will do | 12:08 |
@wiking | fs_: grat thnx | 12:12 |
@wiking | *great | 12:12 |
@wiking | naywhayare: ping | 12:14 |
-!- kahne2012 [~skhan@183.157.160.25] has joined #shogun | 12:17 | |
-!- lambday [67157d54@gateway/web/freenode/ip.103.21.125.84] has quit [Ping timeout: 245 seconds] | 12:21 | |
kahne2012 | hi everyone, I am a new comer :-D | 12:23 |
@wiking | hah | 12:27 |
@wiking | lisitsyn1: i'm on fire today :D | 12:27 |
@wiking | kahne2012: that's great! | 12:27 |
naywhayare | wiking: ping response | 12:29 |
@wiking | naywhayare: ah so here's a new possible joint project :D | 12:30 |
@wiking | naywhayare: ML as SaaS | 12:30 |
@wiking | naywhayare: so basically the idea would be to implement this spec: http://psi.cecs.anu.edu.au/spec and create a wrapper for the ml libraries that join this | 12:31 |
naywhayare | wow, somebody spent a lot of time writing that | 12:32 |
@wiking | naywhayare: yeah and one of the author said he'd be keen to join and mentor this gsoc project | 12:32 |
@wiking | Mark Reid | 12:33 |
@wiking | but i think this should be a similar thing to benchmarking | 12:33 |
@wiking | i.e. most of it shouldn't be coupled to a given ML library | 12:33 |
naywhayare | oh neat, Mark Reid | 12:33 |
naywhayare | he works with a guy also at ANU named Bob Williamson who works with a former labmate who's now a postdoc there | 12:34 |
@wiking | naywhayare: so if u guys would be interested in this project in mlpack then we could do as well like a joint gsoc or something | 12:34 |
naywhayare | I personally don't have very much interest in that project or SaaS in general, but I'm happy to help guide someone who's trying to implement something, assuming that mlpack doesn't need to be accepting contributions for all kinds of SaaS support | 12:34 |
@wiking | naywhayare: nah the idea of course would be to create a general SaaS server and then do the wrapping for the different ML libraries | 12:35 |
@wiking | so imo very small amount of code would go into the ml library itself | 12:35 |
@wiking | if any | 12:35 |
@wiking | same as benchmark | 12:36 |
naywhayare | ok, that is what I figured. like I said, I don't have much interest but I'm happy to help point someone in the right direction while they are working on it, if they are wrapping mlpack | 12:36 |
naywhayare | I don't know who the intended audience of this SaaS system is though | 12:36 |
@wiking | naywhayare: i'd say pure users of ML, that doesn't want to deal with the internals of an ML library | 12:37 |
naywhayare | is Mark Reid going to set this up and run it off a website or something? where maybe a user can interact with it or something? | 12:37 |
@wiking | naywhayare: yeah that'd be the idea i guess | 12:37 |
@wiking | you can setup a service and just interract with a rest api | 12:38 |
@wiking | and have yourself a classifier | 12:38 |
naywhayare | seems interesting enough. so did Mark get in touch with you guys about this, or did you guys just find it and think it was neat? | 12:40 |
@wiking | naywhayare: somebody came to #shogun with this idea based on our docker container | 12:40 |
@wiking | and then Cheng Soon Ong told us about this whole project that Mark is doing | 12:40 |
@wiking | and then Cheng contacted Mark (since they know each other) and Mark told him that he'd be happy to be part of the mentors for this project | 12:41 |
@wiking | naywhayare: ^ the short story ;) | 12:41 |
naywhayare | let me send an email to the guy I know ( http://users.cecs.anu.edu.au/~nmehta/ ) and see what he knows about this project -- he might be pretty familiar | 12:43 |
naywhayare | and he's also familiar with mlpack, so maybe he could help out too, if he's not too busy dodging giant spiders and sharks in australia | 12:43 |
@wiking | naywhayare: hahah canberra is an awefully quite place :) | 12:43 |
@wiking | so there's really not much to do there :) | 12:44 |
@wiking | and no sharks either | 12:44 |
@wiking | they have an artifical lake that you cannot use for anything as it's always having some bacteria thing going on ;P | 12:44 |
naywhayare | hm. I suppose I was misinformed then :) | 12:44 |
@wiking | yep... i lived 1.5 years there | 12:44 |
@wiking | so i know exactly how the place is | 12:44 |
@wiking | ;) | 12:44 |
naywhayare | I need to run... back in about 15 | 12:44 |
@wiking | naywhayare: ok so just think about it and let me know if u wanna share a slot on this or not ... | 12:45 |
lisitsyn1 | wiking: what is the place you find the most exciting? | 12:49 |
abinash_panda | wiking: Hi ! | 12:53 |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 12:54 | |
@wiking | abinash_panda: dont pm me plz | 13:02 |
@wiking | write it here | 13:02 |
abinash_panda | wiking: Ok. I saw your comments on the issue https://github.com/shogun-toolbox/shogun/issues/1913. Being a beginner in shogun I was very confused with the IO, so sorry for the disobedience :) | 13:03 |
abinash_panda | Should I go forward for writing an implementation of CUAIFile or is there any thing more to discuss ? | 13:05 |
@wiking | abinash_panda: well yeah CUAIFile should be a general UAI file parser | 13:07 |
@wiking | and all the file modification, i.e. read/write should be done there | 13:07 |
naywhayare | wiking: what does it mean to share a slot? keep in mind we are relatively new at GSoC :) | 13:08 |
@wiking | i reckon the current CFile api is not going to serve your needs so you'll need to inherite CUAIFile directly from CSGObject... | 13:08 |
@wiking | abinash_panda: but i might be wrong.. | 13:09 |
@wiking | naywhayare: well actually i dont know what's the deal with the benchmarking task atm | 13:09 |
@wiking | naywhayare: thing is that each gsoc project gets a limited number of slots fro google | 13:09 |
@wiking | i.e. it's usually smaller than you'd like to have | 13:09 |
@wiking | so what i thought that with this 2 projects we could share slots | 13:10 |
@wiking | i.e. u take the benchmarking and we take the SaaS | 13:10 |
@wiking | this way we dont waste 2-2 slots for each of those tasks | 13:10 |
naywhayare | ok, that seems reasonable. I had thought we'd each throw a student at the benchmarking task, but I think just one student is fine too | 13:10 |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Ping timeout: 245 seconds] | 13:11 | |
@wiking | naywhayare: u reckon 1 student wouldn't b enough for doing the benchmarking? :) | 13:11 |
abinash_panda | wiking: Ok. But most of the files like CCSVFile, CSVLibSVMFile and CHDF5File inherit from CFile. It is that the CFile API have been changed after that | 13:11 |
@wiking | i mean i dont know.... i think you know it better | 13:11 |
naywhayare | wiking: well, my thought was that each library has different things they want to see; coordinating between libraries on one unified set of goals for the framework might be difficult. but if each library throws a student at it, then each library can take the project in their own direction a bit | 13:12 |
abinash_panda | wiking: I am relatively new to shogun :) Yeah, but surely I would look into it. | 13:12 |
@wiking | abinash_panda: cfile requires u to implement Vector Access Functions and such which i believe you dont have at all in UAI right? | 13:12 |
@wiking | naywhayare: mmmm that sounds good as well | 13:13 |
naywhayare | this doesn't mean that coordination is impossible of course, or that having only one student isn't sufficient, we'll just have to figure out what the student should be doing and define how much of their work is focused on a particular library and how much is focused on the benchmarking system | 13:13 |
naywhayare | (hopefully IRC did not cut off that message... I forget how long the limit is) | 13:13 |
@wiking | naywhayare: nono i've got all that | 13:13 |
@wiking | naywhayare: yeah ok that sounds like a good direction... or even better than i thought | 13:13 |
@wiking | naywhayare: but we'll have to keep in mind that both of us will has less slots than we'd desire :P | 13:14 |
@wiking | abinash_panda: so basically i think since you dont have matrices, vectors, sparse vectors etc. in UAI (or yes?) i think you will just create a new CUAIFile | 13:14 |
@wiking | that represents the content of an UAI file in a structured format with a good api... | 13:15 |
naywhayare | wiking: definitely true. this might be something we can coordinate further when we find out how many good students there are who are interested in the benchmarking project | 13:15 |
abinash_panda | wiking: the Function table in UAI file format is a vector. So, I think that vector access functions might be needed | 13:16 |
@wiking | abinash_panda: ok then you might want to have CFile inheritance after all.. | 13:16 |
@wiking | abinash_panda: it all depends on the file and it's content | 13:17 |
abinash_panda | wiking: Yeah. | 13:17 |
@wiking | abinash_panda: obviously if CFile API is really not something you need you dont have to use that | 13:17 |
@wiking | it's just that the io handling should be in io | 13:17 |
@wiking | as this whole UAI story is Factor Graph independent | 13:17 |
@wiking | it's a file format | 13:17 |
@wiking | and that's all | 13:17 |
@wiking | naywhayare: yeah i guess we'll find this out next month. | 13:18 |
naywhayare | I still don't have a response from the scikit-learn developers yet. that will shed some light on the right way to go, too | 13:18 |
abinash_panda | wiking: Yeah. After the long discussion on the mailing list I totally got the idea about the UAI story :) | 13:18 |
@wiking | abinash_panda: good, then you can refactor your previous PR according to the discussion | 13:19 |
abinash_panda | wiking: Yeah. I would be doing it ASAP. | 13:19 |
abinash_panda | wiking: I am having some exams right now. As soon as I am free I would refactor my PR. | 13:20 |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 13:21 | |
-!- kprah [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 13:34 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has joined #shogun | 13:39 | |
kprah | wiking: dualibqpbmsosvm contains common interface to all three. I have unified libbmrm.cpp, libp3bm.cpp and libppbm.cpp. | 13:40 |
@wiking | kprah: yeah i know about dualibqpbmsosvm... been already working with shogun when it was introduced... but that wasn't the point of that issue | 13:42 |
@wiking | the point was to somehow unify some of the code between libbmrm.cpp, libp3bm.cpp and libppbm.cpp. | 13:42 |
@wiking | and that somehow start with them to define a common optimisation framework... | 13:42 |
-!- kprah_ [67157d4c@gateway/web/freenode/ip.103.21.125.76] has joined #shogun | 13:42 | |
@wiking | kprah: since libbmrm.cpp, libp3bm.cpp and libppbm.cpp is just one big function instead of a structured cpp class.. | 13:43 |
kprah_ | i have made a function, con_bmrm_solver. | 13:43 |
kprah_ | and combined p3bm, bmrm, ppbm | 13:44 |
kprah_ | is that not needed? | 13:44 |
-!- kprah [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [Ping timeout: 245 seconds] | 13:44 | |
@wiking | kprah: its better that u should this either in a pr or like on gist | 13:44 |
kprah_ | ok. | 13:44 |
@wiking | it's not the easiest thing to talk about a code that i have not seen | 13:45 |
@wiking | just heard about in 2 lines | 13:45 |
kprah_ | let me quickly verify the code. | 13:45 |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 13:49 | |
-!- abinash_panda [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has quit [Ping timeout: 245 seconds] | 13:49 | |
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 246 seconds] | 13:55 | |
-!- dhruv13 [~dhruv13@84.64.15.174] has joined #shogun | 14:07 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 14:08 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Ping timeout: 245 seconds] | 14:09 | |
dhruv13 | besser82: hey! i was reading up about the svm light rewrite project... it seems pretty interesting | 14:11 |
@besser82 | dhruv13: Yeah! That would be mine to mentor then ;) | 14:11 |
dhruv13 | besser82: yeah! what do you expect, mathematically? | 14:12 |
dhruv13 | besser82: from the student who does it, i mean | 14:12 |
dhruv13 | besser82: has anyone (better) applied for it? | 14:13 |
@besser82 | dhruv13: basically understanding of the pr_loqo-solver developed by A. Smola ;) | 14:13 |
@besser82 | dhruv13: noone, but xyou yet. ;) | 14:13 |
dhruv13 | besser82: cool! so to convince you, I'll have to read up on the paper by Thorsten Joachims right? | 14:16 |
dhruv13 | besser82: I realize I'm being ambitious here, but I think I can learn enough... | 14:16 |
dhruv13 | besser82: do you mean this? http://www.princeton.edu/~kung/ele571/571-MatLab/571svm/pr_loqo.c | 14:17 |
@besser82 | dhruv13: yes that one :D | 14:18 |
@besser82 | dhruv13: there is modded version of that in shogun as well, which has been extended by sonney2k and Gunnar Rätsch | 14:18 |
@besser82 | dhruv13: I believe in you :D You can doooooooooooo it! :D | 14:19 |
dhruv13 | besser82: yep, went through it! | 14:19 |
@wiking | dhruv13: btw: it's already imported into shogun: src/shogun/lib/external/pr_loqo.h src/shogun/lib/external/pr_loqo.cpp | 14:19 |
dhruv13 | besser82: heh he, thanks!! | 14:19 |
@besser82 | wiking: just mentioned that ;) thx! | 14:19 |
dhruv13 | besser82: cool! I'll start reading up!! | 14:21 |
@besser82 | dhruv13: feel free to contact me for more details :D | 14:21 |
dhruv13 | besser82: sure! | 14:21 |
-!- kahne2012 [~skhan@183.157.160.25] has left #shogun [] | 14:38 | |
-!- kahne2012 [~skhan@183.157.160.25] has joined #shogun | 14:40 | |
-!- kahne2012 [~skhan@183.157.160.25] has quit [Quit: Leaving.] | 14:41 | |
-!- kahne2012 [~skhan@183.157.160.25] has joined #shogun | 14:42 | |
-!- kahne2012 [~skhan@183.157.160.25] has left #shogun [] | 14:42 | |
-!- Saurabh7 [~Saurabh7@115.248.130.148] has joined #shogun | 14:54 | |
-!- cameo54321 [6f4460e2@gateway/web/freenode/ip.111.68.96.226] has joined #shogun | 14:56 | |
-!- abinash_panda [0e8b2997@gateway/web/freenode/ip.14.139.41.151] has joined #shogun | 15:27 | |
kprah_ | wiking: do you reccommend making an interface to liblinear, mosek and libgfs? | 15:31 |
-!- abinash_panda [0e8b2997@gateway/web/freenode/ip.14.139.41.151] has quit [Quit: Page closed] | 15:35 | |
-!- abinash_panda [01261510@gateway/web/freenode/ip.1.38.21.16] has joined #shogun | 15:36 | |
@wiking | kprah_: sorry | 15:38 |
@wiking | kprah_: this pr is just not it | 15:38 |
@wiking | kprah_: what u did is basically copy pasted the contents of the 3 files | 15:39 |
@wiking | and used a big switch | 15:39 |
kprah_ | most of the methods are different in all the solvers? | 15:40 |
kprah_ | I mean the original methods? | 15:40 |
@wiking | mmm | 15:40 |
@wiking | i have a feeling that we dont undertand each other properly :( | 15:41 |
kprah_ | ok | 15:41 |
kprah_ | what should i do then? | 15:41 |
@wiking | well first it would be good that you dont copy paste things... rather try to refactor and create classes | 15:42 |
@wiking | i mean what the whole gsoc project will be about (lp/qp optimization framework) requires a lot of c++ knowledge | 15:42 |
@wiking | and from the PRs and patches you've sent in so far | 15:42 |
@wiking | i couldn't see that you get the idea of refactorization of a code | 15:42 |
@wiking | and especially | 15:43 |
@wiking | do not add files like | 15:43 |
@wiking | src/shogun/structure/DualLibQPBMSOSVM2.h/cpp to the repository | 15:43 |
@wiking | if you have to modify the original file it's alright | 15:43 |
@wiking | modify it | 15:43 |
@wiking | but don't copy and then modify it | 15:43 |
kprah_ | ok. So last chance. Please tell me what is to be done. | 15:45 |
@wiking | kprah_: last chance? | 15:45 |
@wiking | what do you mean by last chance? | 15:45 |
kprah_ | I mean i already have made so many mistakes | 15:45 |
@wiking | well from mistakes one can learn | 15:46 |
@wiking | so that's not a problem at all | 15:46 |
-!- cameo54321 [6f4460e2@gateway/web/freenode/ip.111.68.96.226] has quit [Ping timeout: 245 seconds] | 15:46 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 15:47 | |
kprah_ | ok. So what i think i should do is make a central bundle class, and inherit each solver from that class? | 15:47 |
@wiking | kprah_: yeah that's a good start | 15:48 |
@wiking | and even more | 15:48 |
@wiking | think about what solvers have in common | 15:48 |
@wiking | and then think about defining an abstract optimizer class | 15:48 |
@wiking | that can be extend for example in a bundle solver class | 15:49 |
kprah_ | ok. i'll make it then. | 15:49 |
@wiking | kprah_: BEFORE you actually again write a code + PR | 15:50 |
@wiking | maybe it'd be better to discuss your ideas | 15:50 |
kprah_ | yeah. i won't forget it now. | 15:51 |
kprah_ | let me gat back to you after some time. | 15:51 |
@wiking | kprah_: yeah just write your ideas in the issue | 15:52 |
@wiking | that way we all can have a discussion about that | 15:52 |
kprah_ | ok. | 15:52 |
-!- kprah_ [67157d4c@gateway/web/freenode/ip.103.21.125.76] has quit [Quit: Page closed] | 15:53 | |
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has joined #shogun | 15:57 | |
_flatline_ | ls | 16:11 |
pfm_ | ls -l | 16:12 |
_flatline_ | lol oops | 16:13 |
_flatline_ | I'm having trouble installing the develop branch. I'm getting this issue: CMakeFiles/Makefile2:116: recipe for target 'src/shogun/CMakeFiles/libshogun.dir/all' failed. Is this a dependency thing? | 16:15 |
-!- kahne2012 [~skhan@183.157.160.25] has joined #shogun | 16:16 | |
-!- kahne2012 [~skhan@183.157.160.25] has left #shogun [] | 16:17 | |
-!- dhruv13 [~dhruv13@84.64.15.174] has quit [Ping timeout: 240 seconds] | 16:20 | |
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has left #shogun ["Leaving"] | 16:24 | |
-!- HeikoS [~heiko@134.91.46.136] has joined #shogun | 16:26 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:26 | |
-!- kahne2012 [~skhan@183.157.160.25] has joined #shogun | 16:34 | |
-!- kahne2012 [~skhan@183.157.160.25] has left #shogun [] | 16:35 | |
-!- HeikoS [~heiko@134.91.46.136] has quit [Ping timeout: 240 seconds] | 16:36 | |
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has joined #shogun | 16:40 | |
-!- HeikoS [~heiko@134.91.46.136] has joined #shogun | 16:46 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:46 | |
-!- [1]Saurabh7 [~Saurabh7@115.248.130.148] has joined #shogun | 16:51 | |
-!- Saurabh7 [~Saurabh7@115.248.130.148] has quit [Ping timeout: 264 seconds] | 16:53 | |
-!- [1]Saurabh7 is now known as Saurabh7 | 16:53 | |
kislay | HeikoS, hey hi :) | 17:09 |
kislay | What are your views about the the pca notebook that I updated. | 17:10 |
kislay | I tried to address the issues you told me, but since they were too many...I fear there may be some left | 17:11 |
@HeikoS | kislay: heya | 17:21 |
@HeikoS | kislay: sorry I did not have lots of time yet, but I will try to review it asap | 17:21 |
@HeikoS | kislay: I really liked the new version | 17:21 |
@HeikoS | but will go through it in detail soon | 17:21 |
kislay | HeikoS, sure :) | 17:22 |
kislay | If theres something you want me to do..that some issues or something, please do tell !! | 17:22 |
-!- tomy18 [~alex@p6.eregie.pub.ro] has joined #shogun | 17:23 | |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has joined #shogun | 17:32 | |
-!- HeikoS [~heiko@134.91.46.136] has quit [Ping timeout: 264 seconds] | 17:34 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has joined #shogun | 17:45 | |
-!- tom_ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has joined #shogun | 17:47 | |
tom_ | hi all! | 17:47 |
-!- aroma [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has joined #shogun | 17:47 | |
-!- tom_ is now known as Guest42750 | 17:47 | |
Guest42750 | I ve got a problem with CSVFile | 17:48 |
Guest42750 | when reproducing the code example (from shogun repo): X_train.load(file_train_feats); | 17:49 |
Guest42750 | nber of features loaded is > than the number of lines in my file | 17:50 |
Guest42750 | sorry: the number of vectors is superior by one towads the number of line in the file | 17:51 |
Guest42750 | is there a size limit when we cannot use CSVFile.load() function ? | 17:52 |
@wiking | Guest42750: shoulnd't be... do you maybe have a header in the CSV/ | 17:54 |
@wiking | ? | 17:54 |
Guest42750 | no header! | 17:54 |
Guest42750 | about 17000 lines in the file | 17:54 |
Guest42750 | feature size is 100 | 17:54 |
Guest42750 | ~10MB | 17:55 |
@wiking | Guest42750: ok what does wc -l <csv file you try to load> says? | 17:55 |
@wiking | Guest42750: r u using cpp directly or some modular interface? | 17:55 |
Guest42750 | 17708 test_localacc_t-2.dat | 17:55 |
Guest42750 | cpp | 17:56 |
Guest42750 | and I got X_train.num_cols = 17709 | 17:56 |
@wiking | oh so it's +1 for some reason | 17:56 |
Guest42750 | yes | 17:56 |
@wiking | Guest42750: do you maybe have an extra empty line on the end of the file? | 17:56 |
Guest42750 | cannot fin why | 17:56 |
Guest42750 | no empty line | 17:57 |
@wiking | Guest42750: does the number num_rows match? i.e. the number of features? | 17:57 |
Guest42750 | yes the number of features match exactly | 17:58 |
@wiking | mmmm | 17:58 |
@wiking | that is very strange | 17:58 |
@wiking | Guest42750: can u maybe do a bisect on the file? :) | 17:59 |
@wiking | Guest42750: basically cat <file name>|head -n1000 > first_1000.csv | 17:59 |
@wiking | and see if that really gives you 1000 feature vectors | 17:59 |
Guest42750 | allright doing it right now | 18:00 |
@wiking | and if it's ok with the first 1000 then start changing it to 2000, 3000 etc. | 18:00 |
@wiking | until you start to have that extra 1 line? | 18:00 |
@wiking | Guest42750: of course you can start with a lower number | 18:01 |
@wiking | Guest42750: like the first 10 lines | 18:01 |
@wiking | just to see if that has a difference ... i.e. there's an extra feature vector or not | 18:01 |
-!- tomy18 [~alex@p6.eregie.pub.ro] has quit [Quit: Konversation terminated!] | 18:02 | |
-!- aroma [75ef5e6f@gateway/web/freenode/ip.117.239.94.111] has quit [Ping timeout: 245 seconds] | 18:07 | |
-!- Vinayah [b49531e7@gateway/web/freenode/ip.180.149.49.231] has joined #shogun | 18:09 | |
-!- Vinayah [b49531e7@gateway/web/freenode/ip.180.149.49.231] has quit [Client Quit] | 18:12 | |
-!- _flatline_ [~flatline@2607:f470:22:8:beae:c5ff:fe5a:903f] has quit [Quit: Leaving] | 18:16 | |
Guest42750 | @wiking 16025 was the line which causes the +1 | 18:16 |
Guest42750 | @wiking but this line seems completly normal in the file .. same number of features | 18:17 |
@wiking | Guest42750: can u copy paste me that line and +-1 line around it? | 18:17 |
@wiking | so basically 3 lines | 18:17 |
@wiking | 16024,16025,16026 | 18:18 |
Guest42750 | 0.0163309 0.017412 0.014451 0.0127551 0.0111471 0.0120288 0.0159575 0.0142758 0.0110557 0.0138211 0.012193 0.0206594 0.0182336 0.0169212 0.0157447 0.0163807 0.0194492 0.0180951 0.0156801 0.0177386 0.0165016 0.0192079 0.0179667 0.0168633 0.0174587 0.0203654 0.0190764 0.016803 0.0187386 0.0175722 0.0151148 0.013785 0.0145072 0.0178999 0.0164184 0.0137111 0.0160246 0.0146436 0.0119952 0.0128187 0.016561 0.0149474 0.0119103 0.0145138 | 18:18 |
Guest42750 | didnt work wait | 18:18 |
@wiking | Guest42750: better use pastebin.com | 18:18 |
Guest42750 | http://pastebin.com/U4Qb1AZV | 18:18 |
@wiking | or some similar service | 18:18 |
Guest42750 | here it is | 18:18 |
@wiking | Guest42750: mmm yeah seems totally legit. could you try this: cat <csvfile>|head -n16026|tail -n3 > testfile | 18:19 |
@wiking | and load that 3 lines test file? :) | 18:20 |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 18:21 | |
@wiking | if that loads good then i guess the only option i have is to generate a random csv file with 100 features + 20k lines | 18:21 |
@wiking | see if that loads ok... | 18:21 |
Guest42750 | test_feats->get_num_features()= 66 test_feats->get_num_vectors()= 3 | 18:22 |
@wiking | 66 features? :) | 18:22 |
Guest42750 | yep | 18:22 |
@wiking | that's not right, it should be 100 no? | 18:22 |
Guest42750 | no 66 is right | 18:22 |
Guest42750 | 3 vectors is right too | 18:22 |
@wiking | ah i thought there are 100 features (you've mentioned something about that..._ | 18:22 |
@wiking | ok | 18:22 |
@wiking | then i'll have to create now a random sample | 18:22 |
Guest42750 | what do u mean ? | 18:23 |
@wiking | Guest42750: just a sec i'll give u a file to test | 18:23 |
Guest42750 | I can send you my data as well, if u want | 18:24 |
@wiking | Guest42750: nono it's easier for me like this :) | 18:26 |
Guest42750 | allright :) brb (3min) | 18:26 |
@wiking | Guest42750: http://maeth.com/test_20k.csv.bz2 | 18:27 |
@wiking | MD5 (test_20k.csv.bz2) = 4b42e4f1a92d9a8754ea35bfd5794aa9 | 18:27 |
@wiking | download and bunzip2 this file and try to test with this | 18:27 |
@wiking | Guest42750: this has 20000 vectors with 70 features | 18:27 |
@wiking | see if that loads for you | 18:28 |
@besser82 | wiking: using md5-hash is a bad idea, if the download-app || webserver doesn't preserve timestamp ;) | 18:28 |
@besser82 | wiking: then it will always be different :( | 18:29 |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has joined #shogun | 18:29 | |
@besser82 | wiking: sha512 is a better approach; doesn't rely on timestamps ;) | 18:29 |
@wiking | besser82: it workedforme for the last 15 years... :) | 18:30 |
Guest42750 | @wiking test_feats->get_num_vectors()= 20001 | 18:31 |
@besser82 | wiking: i didn't say it won't work, just pointed out the issue when timestamp is different ;) | 18:31 |
hushell | wiking: need to clarify here, all IO stuffs should be in a IO class? | 18:31 |
@wiking | hushell: yes | 18:31 |
@wiking | besser82: well md5 has bigger issues than timestamp obviously | 18:31 |
@wiking | besser82: but it's goodenough for this matter | 18:31 |
hushell | wiking: but if what I want is more than a standard UAI format? | 18:31 |
@wiking | hushell: extend the UAI format? | 18:31 |
@besser82 | wiking: yes, i know. :D | 18:31 |
hushell | wiking: yeah | 18:31 |
hushell | wiking: we have factor types | 18:32 |
@wiking | hushell: well then just inherit a class from UAIFile and do the extension there | 18:32 |
@wiking | hushell: if that's not enough then create a brand new IO class | 18:32 |
hushell | wiking: their function tables is necessary for our case | 18:32 |
Guest42750 | wiking: feats->get_num_vectors()= 20001; should I recompile lastest version of shogun ? | 18:32 |
hushell | wiking: is not necessary I mean | 18:32 |
@wiking | Guest42750: what's the version do u use? :) | 18:33 |
Guest42750 | wiking: recent one | 18:33 |
hushell | wiking: great! | 18:33 |
@wiking | Guest42750: meaning? | 18:33 |
hushell | abinash_panda: so you have 2 classes to create now :) | 18:33 |
@wiking | hushell: well if it's not necessarry then just start a UAI file implementation that can handle the <preamble> | 18:34 |
hushell | abinash_panda: one for standard UAI, another is a derived class for our purpose | 18:34 |
@wiking | hushell: it very much depends on the actual file | 18:34 |
@wiking | hushell: none the less file creation/parsing should be in IO | 18:34 |
Guest42750 | wiking: 3.2.0 | 18:34 |
@wiking | hushell: decoupled from an implemenation of an algorithm | 18:34 |
@wiking | Guest42750: yeah that should be good | 18:35 |
hushell | wiking: got your point | 18:35 |
@wiking | Guest42750: i guess we have a bug unfortunately in CSVFile then... | 18:35 |
hushell | wiking: thanks for pointing this out | 18:35 |
@wiking | hushell: nw | 18:35 |
Guest42750 | wiking: did you reproduce the "feats->get_num_vectors()= 20001" with your file as well ? | 18:37 |
Guest42750 | wiking: smell like mem leaks ... | 18:37 |
@wiking | Guest42750: haven't tested it sorry | 18:37 |
@wiking | Guest42750: i've made you to do the test :D | 18:37 |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 18:38 | |
Guest42750 | wiking: most of the cpp shogun examples use CSVFile.load() :) | 18:38 |
@wiking | yeah i know | 18:39 |
abinash_panda | hushell: Hi ! | 18:42 |
abinash_panda | hushell: So, should I go for creation of 2 classes or not ? | 18:43 |
@wiking | Guest42750: https://github.com/shogun-toolbox/shogun/issues/1991 | 18:45 |
@wiking | Guest42750: i've just created an issue, let's hope somebody will invetigate+fix soon | 18:45 |
-!- sunil1337 [0e8bd057@gateway/web/freenode/ip.14.139.208.87] has quit [Ping timeout: 245 seconds] | 18:49 | |
@wiking | Guest42750: thnx for reporting the bug | 18:49 |
Guest42750 | wiking: np! | 18:50 |
-!- Guest42750 [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has quit [Quit: Page closed] | 18:51 | |
hushell | abinash_panda: See my updated description | 18:52 |
hushell | abinash_panda: yeah, there will 2 classes | 18:53 |
abinash_panda | hushell: Ok. I got it. | 18:54 |
abinash_panda | hushell: BTW, I am relatively new to shogun and this IO this is very scarry to me. | 18:54 |
abinash_panda | hushell: So, can I use STL for loading and saving instead of CLineReader, CTokenizer, etc. | 18:55 |
@wiking | abinash_panda: what's the problem with linereader? | 18:58 |
abinash_panda | wiking: I have not yet faced any problems. I am currently reading the documentation to get acquainted. | 19:00 |
abinash_panda | wiking: Initially, I have written the save and load method using STL methods. So, I was just asking about it. | 19:00 |
@wiking | abinash_panda: i'd b great if you could use shogun's classes for this | 19:01 |
@wiking | abinash_panda: as for example lineReader has buffering and some other funky things going on | 19:01 |
@wiking | abinash_panda: obviously if the neither of the classes in shogun helps you in your parsing/writing then you should just use whatever you feel you need to use | 19:02 |
abinash_panda | abinash_panda: Yeah sure. I would use CLineReader. It would be interesting to learn | 19:02 |
hushell | abinash_panda: the purpose of that task is to understand factor graphs and other Shogun classes. I'd suggest you to use classes in Shogun to do this. | 19:02 |
@wiking | abinash_panda: but as a rule of thumb try always to use shogun classes if possible | 19:02 |
abinash_panda | wiking: Ok. Currently, I am going through the documentation. Would ping you in case of any doubt :) | 19:03 |
@wiking | abinash_panda: moreover because we indend to add native archive support for file reading/writing (like gziped/bzip2-ed csv etc.) and in this case if you write the parsing here using purely std functions then we won't be able to read archived UAI files | 19:04 |
abinash_panda | wiking: Ah! I see. Then I would be using shogun classes | 19:05 |
@wiking | but then again as said maybe the classes currently available in shogun does not fulfill your needs | 19:06 |
@wiking | so you have to be rational about this... | 19:06 |
abinash_panda | Yeah! I would take that into consideration :) | 19:07 |
@wiking | thnx | 19:07 |
PirosB3 | hey guys | 19:20 |
PirosB3 | hey wiking, I had a chat with Mark Reid, he made the PSI spec for a restful machine learning API. | 19:20 |
PirosB3 | he says he would be happy to mentor my project | 19:20 |
-!- hushell [~hushell@c-50-188-151-130.hsd1.or.comcast.net] has quit [Quit: WeeChat 0.3.7] | 19:21 | |
PirosB3 | can he be my mentor? :) | 19:21 |
@wiking | PirosB3: that all depends. i've written him an email today but i guess he'll answer it later today. it depends on a lot of things. first of all mark ried would be one of your mentors in the project. i guess he would be more like the specs guy... and in this particular project i guess i would be your other mentor... but still it very much depends on a) if all of us in shogun are convinced that this project can work out within gsoc's timeframe b) the pers | 19:26 |
PirosB3 | wiking: oh cool you would be my mentor | 19:27 |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 19:27 | |
@wiking | PirosB3: and maybe somebody from mlpack as well because it would be interesting to integrate not only shogun into this framework but other ml libraries | 19:28 |
@wiking | PirosB3: as said it depends on a lot of factor | 19:28 |
PirosB3 | wiking: I will be chatting with him tomorrow. The idea would be to define some milestones compatible with SoC and then post them out to the community | 19:28 |
PirosB3 | also, given this task, do you suggest any tickets? | 19:28 |
@wiking | PirosB3: the factor you can influence actually all of those is to solve as much as possible entrance tasks so that we see your skills | 19:29 |
@wiking | *out of all of those... | 19:29 |
PirosB3 | ok | 19:29 |
PirosB3 | but you say I should also pitch other libraries? why is that? | 19:30 |
@wiking | PirosB3: well there are entrance tasks that are not related at all to ML but basic bugs that needs to be fixed | 19:30 |
@wiking | PirosB3: well because actually what Mark et.al. sketched up in that specs could be applied to any ML library | 19:30 |
@wiking | PirosB3: most of it should not be bundled to shogun actually | 19:30 |
PirosB3 | wiking: that would be creating another layer of abstraction | 19:31 |
PirosB3 | wiking: do you think I will manage to learn more about ML in general and how to apply it, with this task? | 19:31 |
PirosB3 | would this actually expand my knowledge on how to use ML correctly? | 19:32 |
@wiking | well since you would need to work a lot with ml algorithms it's kind of hard not to learn on the end of this project what is what | 19:39 |
naywhayare | someone said mlpack, here I am | 19:40 |
@wiking | PirosB3: but again your main goal now should be to send in more PRs fixing some pending issues that we can actually quantify somehow your understanding of shogun as well as pick up the way shogun is being developed | 19:40 |
@wiking | naywhayare: u have a highlighter for mlpack? :) | 19:40 |
@wiking | naywhayare: btw i'm just talking with PirosB3 about the SaaS story... | 19:41 |
naywhayare | I do, I'm happy that I did that. I thought we were talking about benchmarking, but actually I should have just continued lurking | 19:41 |
PirosB3 | hi naywhayare | 19:41 |
naywhayare | because I haven't actually looked much into the SaaS ideas | 19:41 |
@wiking | naywhayare: hehehe | 19:41 |
@wiking | naywhayare: nw it's still in an early phase... | 19:41 |
naywhayare | I am glad I read this though because this helps me understand the idea :) | 19:41 |
PirosB3 | naywhayare: have you read my SaaS specs? what do you think of them? | 19:42 |
@wiking | PirosB3: he just said that 'I haven't actually looked much into the SaaS ideas' | 19:43 |
naywhayare | PirosB3: I saw the link to what I thought was a document written by Mark Reid | 19:43 |
PirosB3 | Oh lol! sry | 19:43 |
@wiking | PirosB3: i've written u some feedback | 19:43 |
@wiking | naywhayare: that is really by mark... | 19:43 |
naywhayare | http://psi.cecs.anu.edu.au/spec | 19:43 |
PirosB3 | oh naywhayare you are Cheng? | 19:44 |
naywhayare | not to my knowledge | 19:44 |
@wiking | PirosB3: nop hi is not | 19:44 |
naywhayare | I'm a guy named Ryan | 19:44 |
naywhayare | anyway, if you have a link to a document you wrote, I'll look at it, but I doubt I can give any useful feedback :) | 19:44 |
@wiking | haven't seen cheng around irc ever... ;) | 19:44 |
@wiking | just on shogun WS :) | 19:44 |
PirosB3 | hi Ryan. Sure that would be great https://docs.google.com/document/d/1KISZL6lyeR-RvWEycTKqOunr3gwrA3PUnxZ8hiJYfgw/edit# | 19:45 |
PirosB3 | wiking: so you are Viktor? | 19:46 |
@wiking | PirosB3: yes... and would be great that u enable commenting... | 19:46 |
@wiking | as this way it's kind of awkward to talk about that document | 19:46 |
PirosB3 | wiking: I will add it now. The reason I disabled it is because I didn't want comments to go in different places, and not be unified in one main plac | 19:47 |
@wiking | PirosB3: i don't really understand the last sentence but nevermind | 19:47 |
-!- kprah [67157f4c@gateway/web/freenode/ip.103.21.127.76] has joined #shogun | 19:47 | |
PirosB3 | wiking: thanks for the comments, I'll post back on the doc | 19:48 |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 19:49 | |
-!- kprah [67157f4c@gateway/web/freenode/ip.103.21.127.76] has quit [Client Quit] | 19:50 | |
-!- PirosB3 [~pirosb3@162.219.6.191] has quit [Quit: PirosB3] | 20:04 | |
-!- khalednasr [2945e719@gateway/web/freenode/ip.41.69.231.25] has joined #shogun | 20:19 | |
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 240 seconds] | 20:27 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 20:40 | |
-!- kislay [~kislay@93.114.45.13] has quit [Read error: Connection reset by peer] | 20:49 | |
-!- abinash_panda [01261510@gateway/web/freenode/ip.1.38.21.16] has quit [Ping timeout: 245 seconds] | 20:53 | |
-!- kislay [~kislay@93.114.45.13] has joined #shogun | 21:13 | |
-!- travis-ci [~travis-ci@ec2-54-204-224-198.compute-1.amazonaws.com] has joined #shogun | 21:40 | |
travis-ci | [travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in sunil1337/shogun: http://travis-ci.org/sunil1337/shogun/builds/20632971 | 21:40 |
-!- travis-ci [~travis-ci@ec2-54-204-224-198.compute-1.amazonaws.com] has left #shogun [] | 21:40 | |
-!- lambday [67157f54@gateway/web/freenode/ip.103.21.127.84] has joined #shogun | 21:43 | |
-!- Saurabh7 [~Saurabh7@115.248.130.148] has quit [Ping timeout: 264 seconds] | 21:43 | |
-!- kislay [~kislay@93.114.45.13] has quit [Ping timeout: 241 seconds] | 21:54 | |
-!- sonne|osx [~sonne@f053047161.adsl.alicedsl.de] has joined #shogun | 22:56 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 23:43 | |
shogun-notifier- | shogun: Viktor Gal :develop * d9c535e / src/shogun/evaluation/ContingencyTableEvaluation.cpp: https://github.com/shogun-toolbox/shogun/commit/d9c535e85ed8dc61d537052a9abce200782b87b2 | 23:43 |
shogun-notifier- | shogun: Remove inline in CContingencyTableEvaluation | 23:43 |
shogun-notifier- | shogun: | 23:43 |
shogun-notifier- | shogun: CContingencyTableEvaluation::get_evaluation_direction function was defined as an inline in the implementation. this caused problems by compilation errors of modular interfaces. | 23:43 |
shogun-notifier- | shogun: thanks for the bug report Wenlin Hu! | 23:43 |
-!- sonne|osx [~sonne@f053047161.adsl.alicedsl.de] has quit [Quit: sonne|osx] | 23:48 | |
--- Log closed Thu Mar 13 00:00:36 2014 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!