--- Log opened Mon Jul 18 00:00:11 2011 | ||
--- Day changed Mon Jul 18 2011 | ||
f-x | sonney2k: so its fine to add a %import "Features.i" line in Library.i? | 00:00 |
---|---|---|
f-x | feels like the class separation would be messed up | 00:00 |
@sonney2k | f-x, it certainly is messed up - but not only there | 00:01 |
@sonney2k | I don't see how this can be resolved though | 00:01 |
f-x | sonney2k: one option is separating all the streaming stuff into a separate file - that would not touch the 'base' interfaces like Library, Features etc. but again that wouldn't be very consistent with the overall scheme | 00:03 |
f-x | on the other hand, the central interfaces would be clean | 00:03 |
f-x | we would need to import Streaming.i from a few of these files though | 00:03 |
@sonney2k | yeah and there comes the problem that it is not consitent | 00:04 |
@sonney2k | whatever clean solution we try to do - we should do it in a way that it is the same from the c++ and the swig interface side | 00:04 |
f-x | sonney2k: as a matter of good design, or will it not work otherwise? | 00:05 |
@sonney2k | yes only wrt design | 00:05 |
f-x | i'm thinking i should wait until (if) i can come up with a better way | 00:05 |
f-x | than messing with all the .i files just for these additions | 00:06 |
@sonney2k | I am afraid that this will require a lot of restructuring | 00:06 |
@sonney2k | I mean there is already some mess in there now - and if we resolve it we should really resolve it | 00:06 |
@sonney2k | that would probably mean that one has to split up library into mulitple pieces | 00:07 |
f-x | sonney2k: i vote we resolve it first, and then i can add this stuff | 00:07 |
@sonney2k | f-x, I would say just try if this 'hack' works and we resolve later | 00:07 |
@sonney2k | otherwise we have too many things to resolve at once | 00:08 |
* sonney2k continues the interface transition | 00:08 | |
f-x | sonney2k: well ok.. i'll try to keep it as clean as i can | 00:08 |
blackburn | good design is not about shogun now :D | 00:09 |
@sonney2k | blackburn, just think of a better design - and we will see if it is possible | 00:11 |
@sonney2k | better design == no mutual includes | 00:11 |
blackburn | sonney2k: I don't know how to it better :) | 00:11 |
bettyboo | ^_^ | 00:11 |
@sonney2k | blackburn, f-x I would really wish we could even modularize shogun further | 00:13 |
@sonney2k | such that one could ship e.g. classifiers in one library separately | 00:13 |
@sonney2k | but I don't see it | 00:14 |
f-x | sonney2k: too many dependencies on other shogun features? | 00:14 |
@sonney2k | yeah | 00:14 |
@sonney2k | like the one you have | 00:14 |
@sonney2k | even for shoguns 'Features' | 00:14 |
@sonney2k | they use Distributions to generate features | 00:14 |
@sonney2k | or Kernels can use distances etc etc | 00:15 |
@sonney2k | f-x, how would one resolve that? | 00:24 |
blackburn | put all the files into megashogun folder | 00:25 |
blackburn | sonney2k: drei mal zwei ist sechs | 00:26 |
@sonney2k | koneschno | 00:27 |
blackburn | sonney2k: ich bin 20 jahre alt | 00:27 |
blackburn | wie alt sind sie? | 00:27 |
@sonney2k | tebja sowut sergey | 00:27 |
blackburn | a tebya zyoren | 00:27 |
blackburn | :D | 00:27 |
blackburn | hmm I have an idea how to joke with chris | 00:29 |
blackburn | I will write him a letter in deutsch | 00:29 |
blackburn | sonney2k: умляут или умлаут? | 00:37 |
@sonney2k | hmmhh compilation times are down to 0 | 00:44 |
blackburn | eeeeh? | 00:45 |
@sonney2k | yeah hard to beat | 00:48 |
-!- blackburn1 [~blackburn@188.122.238.99] has joined #shogun | 02:02 | |
-!- blackburn1 [~blackburn@188.122.238.99] has quit [Client Quit] | 02:04 | |
-!- blackburn [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 02:05 | |
CIA-87 | shogun: Soeren Sonnenburg master * rb01925e / (1663 files in 79 dirs): | 02:11 |
CIA-87 | shogun: Restructure repository, such that | 02:11 |
CIA-87 | shogun: - libshogun and libshogunui are merged and in shogun subdir | 02:11 |
CIA-87 | shogun: - all interfaces are found in src/interfaces/<interface> | 02:11 |
@sonney2k | probably the biggest change ever | 02:11 |
CIA-87 | shogun: - static interfaces have a _static suffix - http://bit.ly/ncU4f3 | 02:11 |
@sonney2k | f-x, did you manage to get it to work? | 02:33 |
f-x | sonney2k: i'm finding i need to include lots of *_includes.i files in other interface files | 02:38 |
@sonney2k | f-x, yes that is 'normal' | 02:38 |
f-x | sonney2k: for example | 02:38 |
f-x | i think | 02:38 |
@sonney2k | I know it is a pain in fact | 02:38 |
@sonney2k | but due to that cross-linking there is no other way - swig is not recursively including anything so you have to unroll the include's | 02:39 |
f-x | sonney2k: exactly | 02:39 |
f-x | so is it sensible to make a 'cumulative' include file? | 02:39 |
@sonney2k | I think that would be even worse | 02:40 |
f-x | when i say include 'Features_include.i', i intuitively assume that all #includes needed by Features.i will be satisfied by that call | 02:40 |
@sonney2k | then whenever that file is touched -> everything needs to be rebuilt | 02:40 |
f-x | sonney2k: right.. with these unnecessary includes i have to add, compilation time is going up,up,up | 02:40 |
@sonney2k | but it is like this no? | 02:40 |
f-x | sonney2k: i don't think so | 02:41 |
f-x | we have to include each file included by Features.i again | 02:41 |
@sonney2k | what is missing? | 02:41 |
f-x | now that Features and Library depend on each other, | 02:41 |
f-x | 1. we have to include all of Feature's includes | 02:42 |
f-x | 2. we have to include all of Library's includes | 02:42 |
f-x | too | 02:42 |
@sonney2k | yeah but that is why I have this inlcude Library_includes.i there | 02:42 |
f-x | (whenever we want to import either of their .i files) | 02:42 |
@sonney2k | %include "Library_includes.i" | 02:42 |
@sonney2k | in Features.i | 02:42 |
CIA-87 | shogun: Soeren Sonnenburg master * rccc1ef3 / (16 files in 4 dirs): | 02:59 |
CIA-87 | shogun: fix errors after occurring with doxygen etc caused by the interfaces/ | 02:59 |
CIA-87 | shogun: move - http://bit.ly/niwCwF | 02:59 |
@sonney2k | f-x, I've restructured the repository | 02:59 |
@sonney2k | but left modular where it was | 02:59 |
@sonney2k | so when you have your patch ready - submit it - then I will also relocate the modular dir | 02:59 |
f-x | sonney2k: great job :) | 03:00 |
f-x | and yeah | 03:00 |
f-x | ok | 03:00 |
@sonney2k | if I had inf time I would do much more polishing in this configure / make stuff | 03:00 |
@sonney2k | but the way it is I only made it work - there is 0 beauty in this | 03:01 |
@sonney2k | anyway - need to sleep now cu! | 03:01 |
-!- f-x [~user@117.192.218.159] has quit [Ping timeout: 260 seconds] | 04:24 | |
CIA-87 | shogun: Soeren Sonnenburg master * rff7b19d / src/configure : don't try to tail configure.log if file doesn't exist - http://bit.ly/pLUExP | 08:53 |
CIA-87 | shogun: Soeren Sonnenburg master * r9c679cc / (11 files in 3 dirs): | 08:56 |
CIA-87 | shogun: Merge pull request #207 from frx/python_stream_bad | 08:56 |
CIA-87 | shogun: StreamingFeatures modular interfaces - http://bit.ly/qFfK5I | 08:56 |
CIA-87 | shogun: Soeren Sonnenburg master * r61a56fc / src/Makefile : generated shogun/Makefile depends on Makefile.template - http://bit.ly/pxJmOG | 09:50 |
CIA-87 | shogun: Soeren Sonnenburg master * r300df4a / (6 files in 3 dirs): fix a number of path issues caused by interface transition - http://bit.ly/rfuGS1 | 09:50 |
CIA-87 | shogun: Soeren Sonnenburg master * r5dbeb21 / (75 files in 3 dirs): move modular into interfaces/ - http://bit.ly/nG1tTf | 09:50 |
CIA-87 | shogun: Soeren Sonnenburg master * r118155e / src/configure : fix configure to not generate a.out cruft - http://bit.ly/nTzYGQ | 10:11 |
CIA-87 | shogun: Soeren Sonnenburg master * rebf230a / (src/INSTALL src/README.developer src/configure): | 10:35 |
CIA-87 | shogun: update documentation to new layout (libshogun never needs to be | 10:35 |
CIA-87 | shogun: specified on ./configure --interfaces=xxx from now on) - http://bit.ly/mVpvJc | 10:35 |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 11:44 | |
blackburn | sonney2k: berserk! | 11:45 |
-!- cwidmer [~quassel@connect.tuebingen.mpg.de] has joined #shogun | 12:15 | |
blackburn | strange to meet you, cwidmer ;) | 12:26 |
-!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 12:44 | |
-!- gsomix [~gsomix@85.26.234.128] has joined #shogun | 13:28 | |
-!- heiko [~heiko@main.uni-duisburg.de] has joined #shogun | 13:45 | |
@sonney2k | heiko, hey! | 13:45 |
@sonney2k | heiko, ping me please if you wake up :D | 14:32 |
-!- gsomix [~gsomix@85.26.234.128] has quit [Ping timeout: 255 seconds] | 14:39 | |
@sonney2k | alesis_, any news on the examples? | 14:50 |
-!- gsomix [~gsomix@85.26.235.120] has joined #shogun | 14:51 | |
-!- gsomix [~gsomix@85.26.235.120] has quit [Ping timeout: 255 seconds] | 14:59 | |
-!- gsomix [~gsomix@85.26.234.160] has joined #shogun | 15:11 | |
heiko | sonney2k, hi there! | 15:18 |
heiko | just woke up :D | 15:18 |
heiko | and also some questions: | 15:18 |
heiko | I want to implement computing confidence intervals for the evaluation of a machine | 15:19 |
@sonney2k | heiko, good morning ;-) | 15:19 |
@sonney2k | there are a number of rather drastic changes in shogun now | 15:19 |
@sonney2k | and one more to come | 15:20 |
heiko | for the confidence intervals, I need to borrow some statistics code, and I thought of perhaps putting this into a new class like CStatistics instead of CMathematics ... but to high priority for now, I just worked on this on the weekend a bit | 15:20 |
@sonney2k | so you should probably have a look | 15:20 |
heiko | ok | 15:20 |
heiko | whats happening? | 15:20 |
@sonney2k | we decided to just re-organize the repository | 15:21 |
@sonney2k | not functional changes (letst hope so) | 15:21 |
@sonney2k | which means | 15:21 |
@sonney2k | we only have a directory src/shogun now with libshogun and libshogunui merged | 15:22 |
@sonney2k | then a directory interfaces | 15:22 |
@sonney2k | with all the interfaces beneath | 15:22 |
heiko | ok | 15:22 |
@sonney2k | including the 'modular' directory | 15:22 |
@sonney2k | and renamed all the python,matlab,... static interfaces to python_static, etc | 15:22 |
@sonney2k | when doing configure one now has to say only --interfaces=python_modular | 15:23 |
@sonney2k | if you don't want anything besides libshogun (that will be always built) then --interfaces= | 15:24 |
@sonney2k | that makes it much easier to grasp what shogun is about | 15:24 |
@sonney2k | and new comers don't have to specify libshogun or so ever | 15:24 |
heiko | ok | 15:24 |
@sonney2k | I am now changing all include paths to look like | 15:24 |
@sonney2k | <shogun/features/SimpleFeatures.h> | 15:25 |
@sonney2k | this way I can directly install the header files into usr/include | 15:25 |
heiko | ok | 15:25 |
@sonney2k | and one can build c++ extensions with libshogun + src | 15:25 |
heiko | just fetched current git | 15:25 |
heiko | building ... | 15:26 |
@sonney2k | heiko, things will change in a minute again though... | 15:26 |
@sonney2k | heiko, regarding the statistics thingy | 15:26 |
@sonney2k | I wonder if we should split up library a bit | 15:26 |
@sonney2k | I mean having some math subdir wouldn't hurt... | 15:26 |
heiko | yes, something like this | 15:26 |
heiko | the mathematics file is huge already | 15:27 |
@sonney2k | library just became too big... I mean i/o for example could also be in a different dir | 15:27 |
heiko | computing a students t distribution would make it even larger | 15:27 |
@sonney2k | yeah | 15:27 |
@sonney2k | and the pvalue stuff I recently added does also not make that much sense in there... | 15:27 |
@sonney2k | however, if we do redesign / relocation we should rather do it in the next 2 weeks | 15:29 |
heiko | ok, I will put the new methods in this statistics file then | 15:29 |
@sonney2k | otherwise we won't get any stability in august | 15:29 |
heiko | the file is additionally then for now and is moved in two weeks | 15:29 |
@sonney2k | heiko, I think we should have a math | 15:31 |
@sonney2k | and an io dir in shogun/ | 15:31 |
@sonney2k | then it is already much more digestible | 15:31 |
@sonney2k | two more swig modules then + more fine grained control possible | 15:31 |
heiko | What about the kernel/distance machine stuff? | 15:32 |
heiko | I think this should be done next | 15:32 |
@sonney2k | ? | 15:32 |
heiko | how they store their model | 15:33 |
heiko | indices/copies of feature vectors | 15:33 |
@sonney2k | ahh yes | 15:33 |
@sonney2k | that is true | 15:33 |
@sonney2k | that is another big issue | 15:33 |
@sonney2k | the only other thing I can think of is that it is pretty messy in swig to do the %import / %include stuff | 15:34 |
@sonney2k | due to e.g. kernels being able to use distances | 15:34 |
@sonney2k | or files that emulate streaming from features | 15:34 |
@sonney2k | these cross-links kill the nice separation we had | 15:35 |
@sonney2k | but on the other hand these are damn cool features. | 15:35 |
@sonney2k | no idea how to fix this.... | 15:35 |
CIA-87 | shogun: Soeren Sonnenburg master * r7719b1c / examples/undocumented/python_modular/distribution_ppwm_modular.py : make ppwm example a bit more meaningful - http://bit.ly/pUUoGH | 15:39 |
CIA-87 | shogun: Soeren Sonnenburg master * re364989 / (548 files in 18 dirs): | 15:39 |
CIA-87 | shogun: change include paths to be all relative to /src, e.g. | 15:39 |
CIA-87 | shogun: <shogun/features/SimpleFeatures.i> - http://bit.ly/nHKwVU | 15:39 |
@sonney2k | heiko, any thoughts about the Modelselection vs ModelSelection issue? | 15:39 |
heiko | sorry, a professor just came in :) | 15:40 |
heiko | but back now | 15:40 |
heiko | the name clash: just change it all to ModelSelection? | 15:41 |
heiko | I do no remember why I exactly did this, but I thought i stepped over something similar: CamelCase in libshogun Camelcase in swig | 15:42 |
heiko | sonney2k, should I just go for the kernelmachine change when I am done with the confidence intervals? | 15:44 |
@sonney2k | heiko, btw I set up another mirror http://code.google.com/p/shogun-toolbox/source/browse/ | 15:45 |
@sonney2k | (code.google.com finally support git too :) | 15:45 |
@sonney2k | heiko, yes | 15:45 |
heiko | ah ok cool, google code is nice :) | 15:46 |
heiko | ok, then I will go for that now ... | 15:46 |
heiko | will change the Modelselection names on the fly | 15:46 |
@sonney2k | heiko, regarding the naming it should be consistent at least | 15:47 |
-!- gsomix [~gsomix@85.26.234.160] has quit [Ping timeout: 250 seconds] | 15:47 | |
@sonney2k | ModelSelection? | 15:47 |
heiko | yes true | 15:47 |
heiko | yes, like that more | 15:47 |
@sonney2k | it is a borderline case | 15:52 |
@sonney2k | Preprocessor vs PreProcessor | 15:52 |
@sonney2k | we chose the first one for that | 15:52 |
heiko | ah yes, that was my source of inspiration :) | 15:52 |
bettyboo | ha ha | 15:52 |
@sonney2k | but it is Preprocessor also in C++ | 15:53 |
@sonney2k | so that is different from your approach :D | 15:53 |
heiko | ok then, | 15:53 |
heiko | i confest :) | 15:53 |
heiko | i do not like camel case ;) | 15:53 |
@sonney2k | you will run into a problem though | 15:53 |
@sonney2k | IIRC you have a class called ModelSelection | 15:54 |
@sonney2k | so you will get a naming clash | 15:54 |
@sonney2k | the module name will clash with the class name | 15:54 |
@sonney2k | that is a problem for java_modular | 15:54 |
@sonney2k | ahh and before renaming please git pull --rebase again | 15:55 |
-!- gsomix [~gsomix@85.26.234.160] has joined #shogun | 15:55 | |
@sonney2k | heiko, nice plan for this week | 15:56 |
@sonney2k | just seen your weekly report... | 15:56 |
heiko | thanks :) | 15:56 |
heiko | hope this does not take too long | 15:56 |
heiko | I will borrow some methods from ALGLIB for the students t distribution | 15:57 |
@sonney2k | the kernel/distance thing can be a nightmare ... | 15:57 |
heiko | yes this thing also | 15:57 |
heiko | hopefully not :) | 15:57 |
@sonney2k | heiko, just make sure that we don't get a licensing clash | 15:57 |
heiko | it is gpl 2+ | 15:57 |
heiko | is that ok? | 15:57 |
@sonney2k | yes perfect | 15:59 |
heiko | ok | 15:59 |
heiko | I initially thought this would be some minor stuff | 15:59 |
heiko | but students t needs gammafunction betafunction and some other stuff | 16:00 |
heiko | so there will be like 5 new methods | 16:00 |
heiko | numerical stuff | 16:00 |
@sonney2k | heiko, gammafunction is already in math right | 16:02 |
@sonney2k | lgamma right? | 16:02 |
heiko | oh | 16:02 |
heiko | did not see | 16:02 |
@sonney2k | heiko, well iti s a standard thing in llibm | 16:03 |
@sonney2k | libm | 16:03 |
@sonney2k | heiko, does ALGLIB have its own implementation for that? | 16:04 |
@sonney2k | if so I would be interested in seeing how much these things differ - I mean who has better numerical precision | 16:05 |
heiko | yes, alglib has its own implementation, seems like they pay much attention on numerical correctness | 16:06 |
heiko | they have it documented in the method comments | 16:06 |
@sonney2k | ok then it doesn't hurt to have their method in there too | 16:10 |
@sonney2k | OK things still compile for all interfaces :-)) | 16:11 |
@sonney2k | so heiko feel free to do the rename | 16:11 |
@sonney2k | heiko, ahh btw regarding the naming clash: we had that with module Kernel and class Kernel | 16:12 |
@sonney2k | so I renamed the class to KernelBase | 16:12 |
@sonney2k | (in swig only) | 16:12 |
heiko | already done, currently building to test whether i forgot something | 16:12 |
@sonney2k | just do the same if you run into problems | 16:12 |
@sonney2k | you will need interfaces=python_modular,java_modular to catch all problems | 16:13 |
* sonney2k thinks we need some public *fast* build-bot just for shogun devel's | 16:13 | |
heiko | ok | 16:13 |
* heiko agrees | 16:14 | |
heiko | that would be quite cool | 16:14 |
@sonney2k | too bad that I cannot afford one from GSoC money... | 16:15 |
heiko | wouldnt the TU-Berlin support this? | 16:15 |
@sonney2k | not really a dedicated server ... | 16:16 |
@sonney2k | just some processor like this http://geizhals.at/deutschland/655274 and 32G of memory and some ugly harddrive | 16:19 |
-!- [1]warpy [~warpy@bzq-79-181-19-104.red.bezeqint.net] has joined #shogun | 16:19 | |
heiko | oh, :( | 16:20 |
@sonney2k | heiko, ? | 16:21 |
heiko | the processor | 16:21 |
@sonney2k | yes what about it? | 16:22 |
heiko | not enough for a fast buildbot | 16:22 |
@sonney2k | heiko, why not? | 16:23 |
@sonney2k | 6cores... | 16:23 |
[1]warpy | hey soeren | 16:23 |
heiko | dont you use it for other stuff? | 16:24 |
[1]warpy | did you get a chance to dig into the c# thing | 16:24 |
heiko | the bioinformaticians here have a 250 core cluster :D | 16:24 |
bettyboo | :> | 16:24 |
@sonney2k | heiko, I've seen shogun compile on some 8-16 core machines. it takes <1 minute | 16:25 |
@sonney2k | only harddrive is the speed limit then | 16:25 |
heiko | cool thing :) | 16:25 |
heiko | here, it takes more than five minutes | 16:25 |
heiko | especially the wrap.cxx files take long | 16:26 |
@sonney2k | heiko, but developing just the library is fast ... only swig based interfaces is a pain, but using ccache + --disable-optimizations --enable-debug --disable-doxygen it is ok | 16:26 |
@sonney2k | heiko, yes these are >2MB files | 16:26 |
heiko | yes, thats the way i do it | 16:26 |
@sonney2k | if you only change anything in a .cpp file it doesn't take any time... | 16:26 |
heiko | yes, true | 16:27 |
@sonney2k | so we should probably put more in *.cpp | 16:27 |
@sonney2k | anything that is not timing critical | 16:27 |
@sonney2k | [1]warpy, I thought daniel wanted to have a look | 16:27 |
@sonney2k | [1]warpy, I didn't yet... but there is some good news: sploving will also work on c# at some point in august. | 16:28 |
[1]warpy | great | 16:40 |
[1]warpy | yeah he was planning to work on it | 16:40 |
[1]warpy | but since we are in a different time zone (we and him) | 16:40 |
[1]warpy | your reply probably came at night | 16:41 |
[1]warpy | i am sure he will only touch it next week | 16:41 |
[1]warpy | i realy hope baozeng helps us. its impressive what he is doing | 16:41 |
@sonney2k | he will do lua and ruby first though | 16:44 |
@sonney2k | and surely c# is more difficult than the other two together | 16:44 |
heiko | sonney2k, just got a linker error, because i do not have ant installed | 16:50 |
heiko | this perhaps should be checked ? | 16:50 |
@sonney2k | surprising... I don't like that we need ant anyways. we could have easily done this w/o | 16:51 |
@sonney2k | but hey java guys :) | 16:51 |
heiko | ;) | 16:52 |
bettyboo | :Q | 16:52 |
@sonney2k | we are lucky that they didn't suck in maven | 16:52 |
gsomix | sonney2k, I'm going to integrate C5.0. All questions will be coordinated with blackburn. Ok? | 16:54 |
@sonney2k | gsomix, great :-) | 16:54 |
@sonney2k | looking very much forward to it | 16:54 |
@sonney2k | I discussed with blackburn recently about it so plan would be to first get it in wrapper style and then modify the internals of C5.0 to use shogun's feature representations. | 16:56 |
[1]warpy | why do you think its more difficult ? | 16:56 |
[1]warpy | we have ton some of the word.. | 16:57 |
CIA-87 | shogun: Soeren Sonnenburg master * r8e49447 / src/configure : test whether ant is installed (and disable java if not available) - http://bit.ly/nySl8t | 17:00 |
@sonney2k | [1]warpy, the typemaps are just not useful as of now | 17:00 |
@sonney2k | [1]warpy, however it would help a lot if you ported the examples | 17:00 |
@sonney2k | this would help sploving a lot | 17:01 |
@sonney2k | heiko, ^ fixed the ant detection | 17:01 |
@sonney2k | hope it tells you that java won't work now :) | 17:01 |
[1]warpy | i did port them | 17:03 |
[1]warpy | he might have added more by most are already ported | 17:04 |
[1]warpy | java and c# are almost identical when it comes to easy code | 17:04 |
@sonney2k | [1]warpy, yeah but we don't know yet if these are working | 17:04 |
[1]warpy | you can check out my fork | 17:04 |
heiko | sonney2k, well, I installed ant in the meantime to check my name changes :) | 17:04 |
[1]warpy | i have no idea how to test those without the typemap thing.. | 17:05 |
[1]warpy | so its a chicken and egg for now :) | 17:05 |
@sonney2k | [1]warpy, it is impossible | 17:05 |
@sonney2k | yes | 17:05 |
@sonney2k | heiko, bah | 17:05 |
* sonney2k renames ant | 17:05 | |
heiko | ehm, what? | 17:05 |
@sonney2k | heiko, btw build time for shogun here with full python_modular 1 minute 43 secs | 17:06 |
heiko | thats good :) | 17:06 |
heiko | I will hopefully buy a new computer soon | 17:06 |
@sonney2k | (core 2 duo - nothing too fancy) | 17:06 |
heiko | main problem here is memory | 17:06 |
heiko | its always full, so harddisc is working a lot | 17:07 |
@sonney2k | [1]warpy, eventually we will get there - but it is better that the main work rests not on my shoulders this time. otherwise we won't get anything done | 17:07 |
heiko | just sent pull request for name change | 17:07 |
@sonney2k | heiko, but memory is el cheapo | 17:07 |
@sonney2k | IIRC I just bought 8GB for this notebook for 40 EUR or so | 17:07 |
heiko | true, but my slots are already full, well perhaps I will buy some | 17:08 |
heiko | and throw the old away | 17:08 |
CIA-87 | shogun: Heiko Strathmann master * r9541603 / (8 files in 2 dirs): renamed Modelselection to ModelSelection - http://bit.ly/oDWEhM | 17:08 |
CIA-87 | shogun: Soeren Sonnenburg master * rd67fba1 / (8 files in 2 dirs): | 17:08 |
CIA-87 | shogun: Merge pull request #208 from karlnapf/master | 17:08 |
CIA-87 | shogun: Modelselection ->ModelSelection - http://bit.ly/pPMG0Q | 17:08 |
@sonney2k | heiko, I have 2x 4GB now in here | 17:09 |
@sonney2k | BTW only because 1 memory module died recently I know all this ... | 17:09 |
@sonney2k | heiko, thx for the patch - exactly as I would have done it!!! | 17:09 |
heiko | hehe ok :) 8GB is something i need here | 17:10 |
* heiko checks amazon | 17:10 | |
[1]warpy | i will try to convert all the examples if there are new ones. i understand that the java examples are considered finished ? | 17:10 |
@sonney2k | [1]warpy, more or less... some still have problems but the big majority works | 17:11 |
@sonney2k | [1]warpy, btw send a pull request and I merge them - chances are good that sploving will modify them when he attempts to do the interface | 17:11 |
heiko | sonney2k, where did you buy your memory? | 17:12 |
@sonney2k | heiko, now imagine what a hexacore would do ... compile time probably 30 secs | 17:12 |
@sonney2k | heiko, I checked geizhals.at | 17:13 |
@sonney2k | geizhals.at/de that is | 17:13 |
heiko | sonney2k, as for the multi-core stuff ... It would be really cool if the grid-search model selection would use munltiple threads | 17:13 |
@sonney2k | heiko, surely it would but I am afraid that we have to do a full redesign to support this | 17:14 |
heiko | really? | 17:14 |
@sonney2k | so better postpone for shogun 2.0 | 17:14 |
heiko | basically all iterations of one loop should be parallel | 17:14 |
heiko | but they are completely independent | 17:14 |
@sonney2k | you would need to create different data splits, different models etc | 17:14 |
@sonney2k | but really clone them or? | 17:15 |
heiko | ok yes, true | 17:15 |
@sonney2k | ok for data you could have multiple data objects with different views somehow | 17:15 |
@sonney2k | but that we don't have right now | 17:15 |
heiko | true, ok then, shogun 2.0 :) | 17:15 |
heiko | next gsoc :) | 17:15 |
@sonney2k | if you do that then you can aswell support some fancy distributed model selection thing... | 17:16 |
@sonney2k | I mean distribute to cluster nodes | 17:16 |
heiko | yes, really cool stuff :) | 17:16 |
heiko | completely cool | 17:16 |
@sonney2k | should be possible since every shogun object can be serialized | 17:16 |
@sonney2k | data should probably get some different handling :) | 17:17 |
heiko | mmh 63€ here for 2*2GB | 17:17 |
@sonney2k | heiko, http://geizhals.at/deutschland/607694 ? | 17:18 |
@sonney2k | no idea what type of memory you need - but the above looks rather cheap | 17:18 |
heiko | i got an inspiron notebook from dell. need this here | 17:19 |
heiko | http://www.geizkragen.de/suche/2GB%20-%20PC2-5300%20-%20DDR2%20667MHz%20200pin%20SO-DIMM%20DDR2#tabdiv | 17:19 |
@sonney2k | 19 EUR | 17:19 |
heiko | oh yes | 17:19 |
heiko | geizkragen is worse than geizhals :) | 17:19 |
@sonney2k | don't use geizkragen | 17:20 |
@sonney2k | that's just crap | 17:20 |
@sonney2k | but you are right DDR2 memory is much more expensive | 17:20 |
@sonney2k | anyway back to work | 17:21 |
heiko | yes :) | 17:21 |
[1]warpy | sonney2k, which uni are you at ? | 17:35 |
-!- heiko [~heiko@main.uni-duisburg.de] has quit [Ping timeout: 258 seconds] | 18:08 | |
-!- [1]warpy [~warpy@bzq-79-181-19-104.red.bezeqint.net] has quit [Quit: HydraIRC -> http://www.hydrairc.com <- In tests, 0x09 out of 0x0A l33t h4x0rz prefer it :)] | 18:49 | |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Read error: Connection reset by peer] | 19:27 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 19:27 | |
blackburn | hey revolutionary guys you have revolutioned it pretty much so far | 19:34 |
-!- gsomix [~gsomix@85.26.234.160] has quit [Ping timeout: 259 seconds] | 19:50 | |
-!- gsomix [~gsomix@109.169.235.179] has joined #shogun | 20:08 | |
-!- gsomix [~gsomix@109.169.235.179] has quit [Ping timeout: 252 seconds] | 20:13 | |
-!- gsomix [~gsomix@85.26.234.160] has joined #shogun | 20:26 | |
-!- gsomix [~gsomix@85.26.234.160] has quit [Ping timeout: 276 seconds] | 20:49 | |
@sonney2k | blackburn, ~~~ooohh~~~soviet~~~~russia~~~you~~~make~~~me~~~feeel~~~like~~~reeeeeeevolution!~~~la~~ | 20:54 |
blackburn | :D | 20:54 |
@sonney2k | blackburn, does everything still work for you? | 20:55 |
blackburn | at morning it did | 20:55 |
blackburn | 'll check now | 20:56 |
@sonney2k | then it will probably still work | 20:56 |
@sonney2k | blackburn, so it is time for more breakage^H^H^H^H^H^restructuring | 20:57 |
blackburn | more?? | 20:57 |
@sonney2k | blackburn, yeah I was thinking the Library.i / the lib/ folder got too big | 20:57 |
@sonney2k | so I would suggest to have an io module | 20:58 |
@sonney2k | and a math module | 20:58 |
@sonney2k | and then split up Mathematics depending on what functions are in there | 20:58 |
blackburn | sonney2k: are sub-namespace possible for C++? | 20:58 |
blackburn | shogun::math? | 20:59 |
blackburn | ah nevermind | 20:59 |
blackburn | bad idea | 20:59 |
blackburn | hehe | 20:59 |
@sonney2k | I don't know - I guess so but I am not a big fan of doing it | 20:59 |
@sonney2k | lib/ has now 86 files | 21:00 |
@sonney2k | that is just too much | 21:00 |
@sonney2k | blackburn, btw http://code.google.com/p/shogun-toolbox/ | 21:02 |
@sonney2k | another mirror | 21:03 |
@sonney2k | I added you as committer | 21:03 |
-!- blackburn1 [~blackburn@188.122.238.99] has joined #shogun | 21:03 | |
@sonney2k | blackburn, so just do git remote add google https://code.google.com/p/shogun-toolbox/ | 21:03 |
@sonney2k | aha | 21:03 |
@sonney2k | who is who now blackburn vs blackburn1? | 21:03 |
blackburn1 | sorry I will be disconnected again hehe | 21:04 |
-!- blackburn [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 21:04 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 21:05 | |
blackburn | okay here more stable now | 21:05 |
blackburn | wired hehe | 21:05 |
blackburn | sonney2k: should i push to google too? | 21:05 |
@sonney2k | yes | 21:06 |
blackburn | okay | 21:06 |
blackburn | it is nice they enabled git | 21:06 |
@sonney2k | what was the last message you'v seen | 21:06 |
@sonney2k | ? | 21:06 |
blackburn | (11:00:29 PM) sonney2k: lib/ has now 86 files | 21:06 |
blackburn | (11:00:56 PM) sonney2k: that is just too much | 21:06 |
blackburn | (11:03:46 PM) sonney2k: who is who now blackburn vs blackburn1? | 21:06 |
blackburn | (11:04:07 PM) blackburn1: sorry I will be disconnected again hehe | 21:06 |
@sonney2k | <sonney2k> blackburn, btw http://code.google.com/p/shogun-toolbox/ | 21:07 |
@sonney2k | <sonney2k> another mirror | 21:07 |
@sonney2k | <sonney2k> I added you as committer | 21:07 |
@sonney2k | --> blackburn1 (~blackburn@188.122.238.99) has joined #shogun | 21:07 |
@sonney2k | <sonney2k> blackburn, so just do git remote add google https://code.google.com/p/shogun-toolbox/ | 21:07 |
@sonney2k | <sonney2k> aha | 21:07 |
blackburn | ah okay | 21:07 |
@sonney2k | having 3 mirrors will make backups obsolete :D | 21:08 |
@sonney2k | so just push there too | 21:08 |
blackburn | yeah hehe | 21:08 |
-!- blackburn1 [~blackburn@188.122.238.99] has quit [Ping timeout: 255 seconds] | 21:08 | |
@sonney2k | so what do you think - should we split up lib/ too or are you already happy now? | 21:09 |
@sonney2k | blackburn, ^ | 21:10 |
@sonney2k | ? | 21:10 |
blackburn | don't know | 21:10 |
blackburn | in fact I don't feel embaressed with too large lib/ :D | 21:10 |
@sonney2k | at least - I like the new structure quite a bit | 21:11 |
@sonney2k | blackburn, yeah but io and math would be very logical sub-units | 21:11 |
@sonney2k | we have like 30 i/o classes in there | 21:11 |
blackburn | if you have time - do it | 21:11 |
@sonney2k | and math will contain some statistics and general and loss functions | 21:11 |
@sonney2k | I never have time | 21:12 |
@sonney2k | but it is now or neer | 21:13 |
@sonney2k | never | 21:13 |
blackburn | yeah, now is better ;) | 21:13 |
@sonney2k | blackburn, do you like the new structure or now? | 21:13 |
@sonney2k | you seem a little quiet today... | 21:13 |
blackburn | exactly it is better than it was | 21:13 |
@sonney2k | blackburn, math or mathematics as subfolders? | 21:15 |
@sonney2k | blackburn, disconnected again? | 21:16 |
@sonney2k | seems so | 21:16 |
* sonney2k chooses mathematics | 21:17 | |
-!- gsomix [~gsomix@178.45.37.165] has joined #shogun | 21:17 | |
blackburn | sonney2k: sorry was afk | 21:18 |
blackburn | sonney2k: up to you, both are ok | 21:18 |
@sonney2k | i guess math will have more clash potential | 21:21 |
* sonney2k enters berserk mode again | 21:29 | |
blackburn | sonney2k: how about add an article to machinelearning.ru? should I? | 21:30 |
blackburn | about shogun of course | 21:30 |
@sonney2k | blackburn, when version 1.0 is out and you have the russian docu ready :) | 21:30 |
blackburn | hmm yes | 21:31 |
blackburn | makes more sense | 21:31 |
@sonney2k | blackburn, btw we have to update NEWS file with all the changes | 21:31 |
@sonney2k | I am slowly loosing track... | 21:31 |
blackburn | yes I was talking bout it yesterday | 21:31 |
blackburn | I completely lost it | 21:32 |
@sonney2k | currently it lists nothing | 21:32 |
@sonney2k | blackburn, btw one more thing that is an easy task is to move quite a lot of code from *.h files into *.cpp | 21:34 |
@sonney2k | all finctions that are not time critical | 21:34 |
@sonney2k | that will increase the size of libshogun but decrease compile time / wrapper size | 21:34 |
@sonney2k | things like get_matrix etc | 21:34 |
@sonney2k | of course this won't work for templates... | 21:34 |
blackburn | sonney2k: ehhh? | 21:35 |
blackburn | how will it improve performance? | 21:35 |
@sonney2k | ehh what? | 21:35 |
@sonney2k | I don't understand this question. | 21:36 |
blackburn | (11:34:00 PM) sonney2k: blackburn, btw one more thing that is an easy task is to move quite a lot of code from *.h files into *.cpp | 21:36 |
blackburn | (11:34:09 PM) sonney2k: all finctions that are not time critical | 21:36 |
@sonney2k | yes exactly | 21:36 |
@sonney2k | so all the multiple include stuff in all the interfaces will go away | 21:37 |
blackburn | how it depends on where is the code? | 21:37 |
@sonney2k | -> smaller wrapper | 21:37 |
@sonney2k | -> faster compilation | 21:37 |
blackburn | so better to write all the code in .cpp? | 21:37 |
blackburn | seriously it is surprise to me | 21:37 |
@sonney2k | if one changes sth in the *.cpp files -> no need to recompile the wrappers | 21:37 |
@sonney2k | currently if you touch a .h file the interfaces are regenerated | 21:38 |
@sonney2k | just touch Mathematics.h | 21:38 |
@sonney2k | and you know what I mean | 21:38 |
@sonney2k | touch Mathematics.cpp -> compile time 2 secs | 21:38 |
blackburn | ahh | 21:38 |
blackburn | okay I will move my code from .h to .cpp now ;) | 21:39 |
blackburn | gsomix: join us hehe | 21:41 |
blackburn | it's ok to be gay lets rejoice with the boys | 21:41 |
@sonney2k | blackburn, well only do this for non-time critical stuff | 21:41 |
@sonney2k | or for code that communicates only within libshogun | 21:41 |
blackburn | how will it depends on performance? | 21:42 |
@sonney2k | for other things it can matter | 21:42 |
blackburn | don't understand | 21:42 |
@sonney2k | blackburn, inline'd functions | 21:42 |
@sonney2k | they will just be copy-pasted to the postition where they are needed | 21:42 |
blackburn | well we inline only get_name stuff and so on | 21:42 |
@sonney2k | so instead of having a function ptr, putting variables on stack etc only the code of the inine function will be called | 21:43 |
blackburn | I heard inline is not a silver bullet ;) | 21:43 |
@sonney2k | fighting vampires today? | 21:44 |
blackburn | hmm | 21:44 |
blackburn | google inlines when function is small | 21:44 |
@sonney2k | few lines or performance critical | 21:45 |
gsomix | blackburn, wut? what to do? | 21:45 |
blackburn | http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Inline_Functions | 21:45 |
@sonney2k | blackburn, but look at get_kernel_col in Kernel.h | 21:46 |
@sonney2k | totally unnecessary that this is in *.h | 21:46 |
blackburn | gsomix: you could help us with moving some code from .h to .cpp | 21:46 |
gsomix | ok | 21:46 |
@sonney2k | too bad that templated functions have to stay in *.h | 21:47 |
blackburn | sonney2k: btw update your fork now | 21:47 |
@sonney2k | blackburn, what did you do? | 21:47 |
blackburn | sonney2k: when? | 21:47 |
@sonney2k | you said update your fork now | 21:47 |
blackburn | oh sorry | 21:47 |
blackburn | :D | 21:48 |
blackburn | gsomix: update your fork now | 21:48 |
blackburn | :D | 21:48 |
* sonney2k orders another vodka for blackburn | 21:48 | |
* sonney2k continues to split up lib | 21:49 | |
gsomix | what files to change? | 21:52 |
blackburn | sonney2k: I'm still compile it without ccache | 21:52 |
* gsomix loves google translate. | 21:53 | |
blackburn | I'm too lame to set it up hehe | 21:53 |
@sonney2k | blackburn, too lazy | 21:53 |
blackburn | what are the files to change | 21:53 |
blackburn | english lessons on #shogun hehe | 21:54 |
@sonney2k | blackburn, should arpack be in mathematics? | 21:54 |
@sonney2k | I would guess so or? | 21:54 |
blackburn | yes it is mathematics :D | 21:54 |
@sonney2k | an lapack too | 21:54 |
blackburn | yeap | 21:54 |
gsomix | blackburn, :) | 21:54 |
@sonney2k | then cplex too | 21:55 |
@sonney2k | ls lib/ | 21:55 |
@sonney2k | Array2.h Array.h Cache.h Compressor.cpp DataType.cpp DynamicArray.h DynInt.h Hash.cpp IndirectObject.h List.h memory.h Signal.cpp Time.cpp Trie.cpp v_array.h | 21:55 |
@sonney2k | Array3.h BitString.h common.h Compressor.h DataType.h DynamicObjectArray.h GCArray.h Hash.h __init__.py memory.cpp Set.h Signal.h Time.h Trie.h | 21:55 |
blackburn | where ArrayN is used? | 21:55 |
@sonney2k | what is that __init__.py doing in there?!?! | 21:56 |
@sonney2k | erasing | 21:56 |
blackburn | someone added it? ;) | 21:56 |
@sonney2k | $ ls mathematics/ | 21:58 |
@sonney2k | arpack.cpp arpack.h Cplex.cpp Cplex.h lapack.cpp lapack.h Loss.h Math.cpp Math.h | 21:58 |
@sonney2k | $ ls io/ | 21:58 |
@sonney2k | AsciiFile.cpp HDF5File.cpp LibSVMFile.h SerializableFile.cpp SerializableJsonFile.h ShogunException.cpp StreamingFileFromFeatures.h | 21:58 |
@sonney2k | AsciiFile.h HDF5File.h MemoryMappedFile.h SerializableFile.h SerializableJsonReader00.cpp ShogunException.h StreamingFileFromSimpleFeatures.h | 21:58 |
@sonney2k | BinaryFile.cpp InputParser.h ParseBuffer.h SerializableHdf5File.cpp SerializableJsonReader00.h SimpleFile.h StreamingFileFromSparseFeatures.h | 21:58 |
@sonney2k | BinaryFile.h IOBuffer.cpp SerializableAsciiFile.cpp SerializableHdf5File.h SerializableXmlFile.cpp StreamingAsciiFile.cpp StreamingFileFromStringFeatures.h | 21:58 |
@sonney2k | BinaryStream.h IOBuffer.h SerializableAsciiFile.h SerializableHdf5Reader00.cpp SerializableXmlFile.h StreamingAsciiFile.h StreamingFile.h | 21:58 |
blackburn | no lib/? | 21:58 |
@sonney2k | File.cpp io.cpp SerializableAsciiReader00.cpp SerializableHdf5Reader00.h SerializableXmlReader00.cpp StreamingFile.cpp | 21:58 |
@sonney2k | File.h io.h SerializableAsciiReader00.h SerializableJsonFile.cpp SerializableXmlReader00.h StreamingFileFromFeatures.cpp | 21:58 |
@sonney2k | $ ls lib/ | 21:58 |
@sonney2k | Array2.h Array.h Cache.h Compressor.cpp DataType.cpp DynamicArray.h DynInt.h Hash.cpp IndirectObject.h memory.cpp Set.h Signal.h Time.h Trie.h | 21:58 |
@sonney2k | Array3.h BitString.h common.h Compressor.h DataType.h DynamicObjectArray.h GCArray.h Hash.h List.h memory.h Signal.cpp Time.cpp Trie.cpp v_array.h | 21:58 |
@sonney2k | looks much better IMHO | 21:58 |
blackburn | agree | 21:59 |
@sonney2k | now let me get this to compile | 21:59 |
@sonney2k | and we need 2 new swig modules too... | 21:59 |
* sonney2k sighs | 21:59 | |
blackburn | sonney2k: do we need swig for maths? | 22:00 |
@sonney2k | blackburn, yeah sometimes - for random numbers | 22:00 |
blackburn | bad | 22:00 |
@sonney2k | certainly not for lapack or so | 22:00 |
-!- gsomix [~gsomix@178.45.37.165] has quit [Ping timeout: 240 seconds] | 22:00 | |
@sonney2k | blackburn, but changing math now won't require the whole library module to be recreated | 22:01 |
@sonney2k | well except if we have these circular includes | 22:01 |
* blackburn wonders if someone could use shogun because it have nice arpack wrapper ;) | 22:01 | |
blackburn | sonney2k: http://dl.dropbox.com/u/10139213/shogun/lle.png | 22:02 |
blackburn | *I was playing with lle while it was compiling | 22:03 |
@sonney2k | nice | 22:03 |
blackburn | better than mds | 22:04 |
-!- gsomix [~gsomix@85.26.232.139] has joined #shogun | 22:04 | |
blackburn | I hope we will do something like that for genomic data with cwidmer ;) | 22:04 |
blackburn | sonney2k: when will you commit your changes? I could better wait for you before doing some changes | 22:06 |
@sonney2k | this will take me >1 h I think | 22:07 |
blackburn | okay, will implement ltsa now hehe | 22:08 |
bettyboo | strange | 22:08 |
-!- genix [~gsomix@85.26.235.146] has joined #shogun | 22:17 | |
-!- gsomix [~gsomix@85.26.232.139] has quit [Ping timeout: 255 seconds] | 22:18 | |
-!- genix is now known as gsomix | 22:19 | |
@sonney2k | blackburn, is gsomix in the same wlan like you? | 22:19 |
blackburn | sonney2k: no, we are pretty far away from each | 22:20 |
blackburn | of us | 22:20 |
@sonney2k | other | 22:20 |
blackburn | aha | 22:20 |
@sonney2k | you meant - ok so how far? | 22:20 |
blackburn | may be even more than 200 km | 22:20 |
@sonney2k | wow - how do you know each other then? | 22:20 |
blackburn | well I lost connection because my wi-fi access point is acting very bad sometimes | 22:20 |
blackburn | we study in one university in samara | 22:21 |
blackburn | but now we both at home | 22:21 |
@sonney2k | blackburn, I recommend one of those http://en.wikipedia.org/wiki/Linksys_WRT54G_series | 22:21 |
@sonney2k | rock stable with e.g. dd-wrt | 22:21 |
blackburn | I have no idea why my wi-fi lost connection on high load | 22:22 |
blackburn | not notebook issues - it was working before | 22:22 |
@sonney2k | hmmhh it did compile?! | 22:23 |
@sonney2k | shall I commit? | 22:23 |
blackburn | why not? | 22:23 |
@sonney2k | because I modified every single file? | 22:23 |
-!- gsomix [~gsomix@85.26.235.146] has quit [Ping timeout: 255 seconds] | 22:24 | |
blackburn | sonney2k: ^ he uses 3g hehe | 22:24 |
@sonney2k | I see | 22:24 |
blackburn | 2514roubles for Linksys WRT54G | 22:25 |
blackburn | will buy later | 22:25 |
blackburn | not much more than dlink hardware | 22:25 |
CIA-87 | shogun: Soeren Sonnenburg master * rbf5a3a3 / .gitignore : | 22:26 |
CIA-87 | shogun: remove shogun from ignored dirs since that is now the main shogun aka | 22:26 |
CIA-87 | shogun: libshogun dir - http://bit.ly/r4mPpC | 22:26 |
CIA-87 | shogun: Soeren Sonnenburg master * r245cf3e / src/Makefile.template : | 22:26 |
CIA-87 | shogun: remove sed -i operation on shogun dir - no longer necessary with new | 22:26 |
CIA-87 | shogun: directory layout - http://bit.ly/oW5PzX | 22:26 |
CIA-87 | shogun: Soeren Sonnenburg master * r6569436 / (369 files in 31 dirs): split up lib/ into io/ lib/ mathematics/ for clarity - http://bit.ly/pAouA2 | 22:26 |
@sonney2k | so committed | 22:26 |
blackburn | now we know what is the file count in shogun | 22:26 |
@sonney2k | I didn't do the change in python modular now | 22:26 |
@sonney2k | I guess I should do it right away | 22:27 |
@sonney2k | blackburn, but please try to compile first | 22:27 |
blackburn | w/o python? | 22:27 |
blackburn | sonney2k: distclean tries to remove shogun | 22:28 |
@sonney2k | great | 22:28 |
blackburn | my luck he can't | 22:28 |
-!- gsomix [~gsomix@85.26.233.192] has joined #shogun | 22:28 | |
@sonney2k | fixed | 22:29 |
CIA-87 | shogun: Soeren Sonnenburg master * r85442d2 / src/Makefile : don't try to remove shogun on distclean - http://bit.ly/nccMxO | 22:29 |
blackburn | sonney2k: does git status show you that ../data is modified? | 22:31 |
@sonney2k | no | 22:32 |
@sonney2k | did yo add sth? | 22:32 |
blackburn | week ago or os | 22:32 |
blackburn | so | 22:32 |
blackburn | sonney2k: btw you intended to *_modular -> *? | 22:33 |
blackburn | compiling.. | 22:34 |
@sonney2k | blackburn, yes - but after consideration I think it is more clear now | 22:34 |
blackburn | sonney2k: okay too much errors I can copy-paste | 22:36 |
@sonney2k | did you git clean dfx before? | 22:37 |
blackburn | git clean | 22:37 |
@sonney2k | OK I've split up Library... | 22:39 |
@sonney2k | now we have Mathematics.i and IO.i too | 22:39 |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 22:41 | |
blackburn | libshogun compiled ok | 22:43 |
-!- blackburn [~blackburn@188.122.238.99] has quit [Quit: Leaving.] | 22:43 | |
-!- blackburn [~blackburn@188.122.238.99] has joined #shogun | 22:43 | |
blackburn | haha I got disconnected even wired | 22:43 |
bettyboo | 8) | 22:43 |
-!- in3xes_ [~in3xes@180.149.49.227] has quit [Client Quit] | 22:44 | |
blackburn | sonney2k: I have K parameter in LLE, how can we add parameter search for heiko's fancy modelselection? | 22:46 |
blackburn | similar for isomap | 22:46 |
blackburn | I can provide reconstruction error by option | 22:46 |
@sonney2k | blackburn, that doesn't work currently | 22:47 |
@sonney2k | it only works for 'machine' derived classes | 22:47 |
@sonney2k | that have a train() and apply() function | 22:47 |
blackburn | hm okay | 22:47 |
@sonney2k | you could probably write a machine for this | 22:49 |
blackburn | yes maybe makes sense | 22:50 |
gsomix | are you need me? i was planning to go sleep. | 22:52 |
blackburn | heh | 22:53 |
blackburn | gsomix: do as you was planning ;) | 22:53 |
blackburn | sonney2k: okay compile | 22:56 |
blackburn | d | 22:56 |
CIA-87 | shogun: Soeren Sonnenburg master * re70754f / (17 files in 2 dirs): split up modular Library interface into Library, Mathematics and IO - http://bit.ly/oE1zMv | 22:57 |
@sonney2k | so finally | 22:57 |
@sonney2k | I guess you will have to update some examples still | 22:57 |
@sonney2k | but it seems to work just fine... | 22:58 |
-!- gsomix [~gsomix@85.26.233.192] has quit [Quit: Ухожу я от вас (xchat 2.4.5 или старше)] | 22:59 | |
@sonney2k | blackburn, I am trying to remove my BoolSparseMatrix) SGSparseMatrix definitions from SGBase.i to Library.i again and then just use imports in SGBase | 22:59 |
@sonney2k | (now that library is small) | 23:00 |
blackburn | I didn't realize what you just said | 23:00 |
blackburn | :D | 23:00 |
@sonney2k | 91 C | 23:07 |
@sonney2k | CPU temperature | 23:07 |
blackburn | I hope it is not notebook? | 23:07 |
@sonney2k | it is | 23:07 |
@sonney2k | macbook pro | 23:07 |
@sonney2k | too hot | 23:07 |
blackburn | yay | 23:07 |
@sonney2k | fans running at 6200 RPM | 23:07 |
@sonney2k | like sitting in a plane | 23:07 |
@sonney2k | unbelievable | 23:08 |
@sonney2k | it still compiles and examples run?! | 23:08 |
@sonney2k | am I doing sth wrong? | 23:09 |
blackburn | make[1]: *** No rule to make target `IO_doxygen.i', needed by `IO_wrap.cxx'. Stop. | 23:09 |
blackburn | I got that thing again | 23:09 |
CIA-87 | shogun: Soeren Sonnenburg master * r2514157 / (13 files in 4 dirs): move data types back to library - http://bit.ly/nalMPD | 23:17 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 23:19 | |
@sonney2k | blackburn, that should fix your problem | 23:20 |
CIA-87 | shogun: Soeren Sonnenburg master * r4c71689 / src/.generate_link_dependencies.py : mention doxy files in link dependency generator - http://bit.ly/rn62Kq | 23:20 |
* blackburn trying to write a paper about IRLM for LLE hehe | 23:23 | |
@sonney2k | blackburn, do we need a Machine module? | 23:24 |
blackburn | in swig? | 23:24 |
@sonney2k | in modular if's | 23:24 |
blackburn | I don't think so, where do we use it? | 23:24 |
@sonney2k | OK one could only use it for trained kernelmachines or so | 23:26 |
blackburn | shit I can't translate 'manifold learning' to russian ahaha | 23:26 |
blackburn | sonney2k: not fixed | 23:31 |
@sonney2k | sure it is | 23:31 |
@sonney2k | I just rebuilt everything... | 23:31 |
blackburn | okay 'll try again | 23:32 |
@sonney2k | dont' forget the git clean -dfx | 23:32 |
@sonney2k | hmmhh naming clash - IO (the module) vs IO (the class) | 23:34 |
@sonney2k | blackburn, suggestions for a better name? | 23:34 |
blackburn | some Logger hehe | 23:36 |
blackburn | ShogunIO | 23:36 |
blackburn | sonney2k: did git clean | 23:38 |
blackburn | make distclean | 23:38 |
@sonney2k | blackburn, git clean -dfx ? | 23:38 |
blackburn | make[1]: *** No rule to make target `IO_doxygen.i', needed by `IO_wrap.cxx'. Stop. | 23:38 |
blackburn | yes | 23:38 |
@sonney2k | but you have a IO_doxygen.i in interfaces/modular/ ? | 23:39 |
blackburn | I have olny IO.doxy | 23:40 |
@sonney2k | err sorry IO.doxy | 23:40 |
-!- alesis_ [~alesis@188.74.87.206] has quit [Ping timeout: 252 seconds] | 23:42 | |
@sonney2k | blackburn, and if you grep IO_ .generate*.py | 23:42 |
@sonney2k | you have it in there or not? | 23:43 |
-!- alesis_ [~alesis@188.74.87.206] has joined #shogun | 23:43 | |
@sonney2k | renaming to SGIO | 23:44 |
blackburn | grep is slow for some reason | 23:44 |
blackburn | ok | 23:45 |
blackburn | it compiled | 23:45 |
blackburn | sorry | 23:45 |
blackburn | sonney2k: I guess I did something wrong | 23:45 |
@sonney2k | ?! | 23:45 |
@sonney2k | heisenbug? | 23:45 |
blackburn | no idea | 23:46 |
--- Log closed Tue Jul 19 00:00:39 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!