--- Log opened Fri Jul 29 00:00:46 2011 | ||
@sonney2k | blackburn, hmmhh I just checked out some hosting solutions for a potential build bot | 00:02 |
---|---|---|
serialhex | sonney2k: i'm working on getting shogun to compile correctly with different versions of ruby, then i will work on making sure the examples work. | 00:02 |
@sonney2k | 60 € / month for a decent quad core with 4G memory | 00:03 |
@sonney2k | should be sufficient I would say | 00:03 |
@sonney2k | 60 EUR | 00:03 |
serialhex | though right now i can't get shogun to work at _all_ with my system set-up, though i have narray & ruby (both 1.8 & 1.9 installed) | 00:03 |
@sonney2k | serialhex, I meant - how do I even run an example? | 00:03 |
blackburn | sonney2k: are you sure we need it? | 00:03 |
serialhex | oh, it should be: ruby example.rb | 00:03 |
@sonney2k | blackburn, I think it is better than running things here on my home machine | 00:03 |
blackburn | which things? | 00:04 |
@sonney2k | at least that would mean faster webserver / other devels could login to build... | 00:04 |
@sonney2k | blackburn, build service | 00:04 |
blackburn | I don't understand.. what is build service? | 00:04 |
@sonney2k | serialhex, yes but when I run it from the examples dir it cannot find modshogun - how do I specify the ruby include path to fix that? | 00:05 |
serialhex | i think sploving got classifier_libsvm_minimal_modular.rb running fine... but idk about any others... | 00:05 |
@sonney2k | blackburn, just building | 00:05 |
serialhex | ahh, ok i see | 00:05 |
@sonney2k | serialhex, I think the others work to - I am just clueless how to run them | 00:05 |
blackburn | well for me it is ok to compile here.. | 00:06 |
serialhex | btw: i *love* physical books... dead trees rock! | 00:07 |
@sonney2k | blackburn, yes sure, but we need automated build tests | 00:09 |
@sonney2k | blackburn, or did you know that static interfaces where all broken for some time? | 00:09 |
@sonney2k | and blackburn we need to run this on osx / cygwin too | 00:11 |
blackburn | hmm I see | 00:11 |
serialhex | ok sonney2k, this is the pain-in-the-ass-until-we-get-ruby_modular-stuff-installed way to do things: ruby -I../../../src/interfaces/ruby_modular example.rb #(that is a capital i) | 00:11 |
serialhex | that will prepend the ruby_modular dir to ruby's $LOAD_PATH and should allow you to run the example from the examples dir | 00:12 |
serialhex | now - off to eat | 00:12 |
@sonney2k | thanks | 00:13 |
@sonney2k | *yawn* | 00:42 |
@sonney2k | nite folks | 00:42 |
blackburn | see you | 00:42 |
-!- blackburn [~blackburn@109.226.104.206] has quit [Quit: Leaving.] | 01:43 | |
-!- in3xes [~in3xes@180.149.49.230] has quit [Quit: Leaving] | 04:57 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 05:58 | |
CIA-87 | shogun: Soeren Sonnenburg master * r107276d / examples/undocumented/ruby_modular/check.sh : add a ruby examples check script - https://github.com/shogun-toolbox/shogun/commit/107276dba7c6f4cb8713b6e81a887f6f8dbada1d | 07:43 |
CIA-87 | shogun: Soeren Sonnenburg master * r620037e / examples/undocumented/ruby_modular/check.sh : add single file test - https://github.com/shogun-toolbox/shogun/commit/620037ed381d27a99b8dc2ce304b85fe58a0159d | 07:43 |
CIA-87 | shogun: Soeren Sonnenburg master * r01b3fc3 / src/NEWS : Merge branch 'master' of github.com:shogun-toolbox/shogun - https://github.com/shogun-toolbox/shogun/commit/01b3fc30fe6a5a4926c0f750177efe8b7100610a | 07:43 |
-!- sploving1 [~sploving@124.16.139.134] has joined #shogun | 09:27 | |
@sonney2k | sploving1, how is it going? | 10:24 |
sploving1 | sonney2k, not work yet | 10:27 |
-!- heiko [~heiko@134.91.54.89] has joined #shogun | 10:40 | |
@sonney2k | sploving1, I have an idea - maybe you *for testing* just add a function set_foo(SGMatrix<float64_t> m) { printf("set\n"); } to Label.h | 10:41 |
@sonney2k | sploving1, or add a constructor to some other class that takes a SGMatrix... | 10:41 |
@sonney2k | moin heiko | 10:41 |
@sonney2k | heiko, you did allocations on stack - no chance the malloc trace stuff couldn detect that... | 10:42 |
heiko | moin | 10:42 |
heiko | oh | 10:42 |
heiko | where did i do that? | 10:43 |
@sonney2k | heiko, in your example | 10:43 |
heiko | oh | 10:43 |
heiko | i see | 10:43 |
@sonney2k | I 'fixed' that | 10:43 |
heiko | thats why it showed allocated blocks | 10:43 |
heiko | damn, could have thought about that | 10:44 |
@sonney2k | heiko, yeah - I mean there was no SG_UNREF for certain objects so ... | 10:44 |
heiko | huh, but valgrind said it was clean | 10:45 |
heiko | i will have a look | 10:45 |
sploving1 | sonney2k, minimal_test.cs in c#dir , how to change it to use modshogun? | 10:46 |
CIA-87 | shogun: Soeren Sonnenburg master * r89bf991 / src/shogun/lib/DataType.h : introduce SGCachedVector prototype - https://github.com/shogun-toolbox/shogun/commit/89bf991a142d09d55b2e99e171e3e655e0f90e90 | 10:46 |
@sonney2k | heiko, btw ^ | 10:46 |
@sonney2k | have a look at my cached vector attempt | 10:46 |
@sonney2k | maybe it works for simplefeautres | 10:46 |
@sonney2k | heiko, feel free to work on this further! | 10:46 |
heiko | ok, | 10:46 |
sploving1 | how let it find modshogun, in c#? | 10:47 |
heiko | sonney2k, you only did change the grid searh linear example, right? | 10:47 |
@sonney2k | yes and more | 10:47 |
@sonney2k | your string slen patch caused build failures all over... | 10:47 |
heiko | oh no :( | 10:47 |
@sonney2k | sploving1, does it need to be modified even? | 10:48 |
heiko | sorry for that, swig typemaps? | 10:49 |
@sonney2k | heiko, typemaps and static interfaces and examples | 10:49 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Remote host closed the connection] | 10:49 | |
sploving1 | sonney2k, yeap. I donot know how to let find the lib modshogun.so. I googled csharp libraray path, but no good results | 10:50 |
@sonney2k | sploving1, just set the LD_LIBRARY_PATH IIRC | 10:51 |
sploving1 | then whant does the Library.init_shogun_with_defaults(); | 10:51 |
sploving1 | GaussianKernel k = new GaussianKernel(); | 10:51 |
sploving1 | 10:51 | |
sploving1 | this is what you write in the minimal_test.cs | 10:52 |
@sonney2k | the same like in java | 10:52 |
sploving1 | you mean deal with c# exactly like java? | 10:52 |
sploving1 | javac path and java path? | 10:53 |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 10:53 | |
@sonney2k | sploving1, ? | 10:53 |
@sonney2k | of course not | 10:53 |
sploving1 | there is no c# expert:( | 10:55 |
sploving1 | here | 10:55 |
-!- blackburn [~blackburn@109.226.104.206] has joined #shogun | 11:08 | |
-!- sploving1 [~sploving@124.16.139.134] has left #shogun [] | 11:13 | |
-!- [1]warpy [~warpy@bzq-79-181-25-215.red.bezeqint.net] has joined #shogun | 11:45 | |
[1]warpy | hey guys, anyone here ? | 11:48 |
@bettyboo | hug | 11:48 |
heiko | hi | 11:53 |
heiko | sonney2, just found this: int32_t *ClList = (int32_t*) calloc(XSize, sizeof(int32_t)); | 11:53 |
heiko | in KMeans, i guess this should be replaced by SG_CALLOC, right? | 11:53 |
@sonney2k | heiko, yes | 12:12 |
-!- gsomix [~gsomix@88.200.214.229] has joined #shogun | 12:12 | |
gsomix | hi all | 12:12 |
@bettyboo | hiho | 12:12 |
@sonney2k | [1]warpy, please ask here | 12:12 |
heiko | sonney2, just thought about distance machines could store their model. I think they just have to completely copy the feature data of their distance variable | 12:16 |
@sonney2k | heiko, ? | 12:17 |
@sonney2k | isn't it the same like kernelmachine? | 12:17 |
heiko | kernelmachine only has to store its SV | 12:18 |
@sonney2k | distance machine too or? | 12:18 |
heiko | kernel machine only has one variable: distance | 12:19 |
@sonney2k | ok KNN is an exception ... | 12:19 |
@sonney2k | heiko, did you mean distance machine? | 12:19 |
* sonney2k is getting confused | 12:19 | |
heiko | yes | 12:19 |
heiko | i was currently working on KNN | 12:19 |
heiko | in apply, the method distances_lhs() is called which compute distances to ALL feature vectors | 12:20 |
heiko | so all of them have to be copied | 12:20 |
@sonney2k | heiko, yes for KNN it is true | 12:20 |
@sonney2k | infact KNN has no training procedure | 12:20 |
heiko | true | 12:20 |
heiko | mmh | 12:21 |
heiko | so store_model does not have to do anything | 12:21 |
@sonney2k | well it needs the training data when testing | 12:22 |
heiko | ok then | 12:22 |
@sonney2k | so I think it is correct to assume that it has to copy all training data | 12:22 |
heiko | KMeans does not even have an apply method, | 12:22 |
heiko | so no use in model selection anyway | 12:22 |
heiko | because it is called | 12:22 |
heiko | same for hierachical | 12:23 |
@sonney2k | heiko, well one could define an apply method in k-means that just assigns the nearest cluster center | 12:23 |
@sonney2k | same for other clustering methods | 12:23 |
heiko | ok | 12:23 |
heiko | but in both KMeans, no model storage is needed since radius and distances are stored in new arrays anyways | 12:24 |
heiko | so DistanceMachine will just not do nothing in store_model_features and KNN will overwrite it and store all feature data | 12:25 |
@sonney2k | heiko, makes sense... maybe more general would be to do distance machine in the same way as kernel machine. | 12:26 |
@sonney2k | on the other hand we don't have such machines yet... | 12:26 |
heiko | yes, and also there is no variable that saves indices of needed features | 12:27 |
@sonney2k | so ok then just do it for KNN | 12:27 |
heiko | sonney2k, just had another thought about kmeans | 12:32 |
@sonney2k | which is? | 12:32 |
heiko | it saves centers for clusters in a variable | 12:32 |
heiko | why not store these in the underlying distance variable | 12:32 |
heiko | then in apply, the distance_lhs method could be used | 12:32 |
heiko | if i do this now, i have to compute distance to all cluster centers by hand | 12:33 |
-!- gsomix [~gsomix@88.200.214.229] has quit [Ping timeout: 252 seconds] | 12:33 | |
@sonney2k | heiko, you mean just create a new feature object with centers? | 12:34 |
CIA-87 | shogun: Heiko Strathmann master * rdba50f2 / (3 files): | 12:35 |
CIA-87 | shogun: -made search space smaller | 12:35 |
CIA-87 | shogun: -removed stuff from stacdk to heap to be able to trace memory blocks | 12:35 |
CIA-87 | shogun: -cleaned up comments - https://github.com/shogun-toolbox/shogun/commit/dba50f2be70b444ea53778284c06ad878bcff047 | 12:35 |
CIA-87 | shogun: Soeren Sonnenburg master * r3ba5d3b / (3 files): | 12:35 |
CIA-87 | shogun: Merge pull request #246 from karlnapf/master | 12:35 |
CIA-87 | shogun: example corrections - https://github.com/shogun-toolbox/shogun/commit/3ba5d3b7fd505a444d5d8bdb6dd3595e5aa79aa2 | 12:35 |
heiko | sonney2k, yes like this | 12:41 |
heiko | what do you think? | 12:41 |
heiko | then the code of DistanceMachine could be reused in apply | 12:43 |
@sonney2k | heiko, yes i like it | 12:44 |
-!- gsomix [~gsomix@109.169.238.89] has joined #shogun | 12:46 | |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 12:51 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 12:55 | |
-!- in3xes__ [~in3xes@180.149.49.227] has joined #shogun | 13:00 | |
-!- in3xes__ is now known as in3xes | 13:01 | |
-!- in3xes_ [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 13:03 | |
[1]warpy | hey sonney2k, when you have 1600 vectors, each has around 2mb of data, what is the best way to run a learning process on it. is it to load it all into memory, or is there some kind of trick that you used in shogun ? | 13:21 |
@sonney2k | [1]warpy, I would use a kernel machine and precompute the kernel matrix | 13:28 |
CIA-87 | shogun: Soeren Sonnenburg master * r2d44753 / (2 files in 2 dirs): | 13:30 |
CIA-87 | shogun: fix compile csharp errors | 13:30 |
CIA-87 | shogun: - made read_vector* variables protected such that there is no get_/set_ | 13:30 |
CIA-87 | shogun: naming clash | 13:30 |
CIA-87 | shogun: - use same signature for get_num_vectors() in attributefeatures to fix warning - https://github.com/shogun-toolbox/shogun/commit/2d44753eb5506eb6415c5dfee71e082754cc83fb | 13:30 |
CIA-87 | shogun: Soeren Sonnenburg master * r035f3cb / examples/undocumented/csharp_modular/minimal_test.cs : fix minimal_test example to work with modshogun - https://github.com/shogun-toolbox/shogun/commit/035f3cbd443ae6eba756fb7cc609305ad90be90c | 13:30 |
CIA-87 | shogun: Soeren Sonnenburg master * r1510d31 / (3 files): Merge branch 'master' of github.com:shogun-toolbox/shogun - https://github.com/shogun-toolbox/shogun/commit/1510d315efcc4018dcbf47838a561ddc6a8d77cc | 13:30 |
@sonney2k | heiko, in your machine train transition - did you make the train_machine function private/protected? | 13:31 |
heiko | yes | 13:32 |
CIA-87 | shogun: Soeren Sonnenburg master * r37e9718 / (30 files in 3 dirs): | 13:32 |
CIA-87 | shogun: Merge pull request #247 from karlnapf/master | 13:32 |
CIA-87 | shogun: completely applied renaming train -> train_machine (was only done for KernelMachines before) - https://github.com/shogun-toolbox/shogun/commit/37e9718706668604cdda9d1f69a20ca0ee3a6ba0 | 13:32 |
@sonney2k | yes | 13:32 |
heiko | yes? | 13:32 |
@sonney2k | yes! | 13:33 |
heiko | :D | 13:33 |
@sonney2k | heiko, I hope you are now always compiling with trace mallocs :) | 13:36 |
heiko | yes i do | 13:36 |
heiko | why? | 13:36 |
heiko | but my examples become very slow this way | 13:36 |
heiko | since there are ~0.5mio allocs :) | 13:36 |
heiko | therefore i made them smaller so that make-tests runs in a reasonable time | 13:36 |
[1]warpy | sonney2k, is this something that is done in shogun with a couple of commands already ? | 13:41 |
@sonney2k | heiko, yeah your examples should be *really* small anyways - these will be run on some poor arm based machines too... | 13:42 |
heiko | true :) | 13:42 |
@sonney2k | [1]warpy, if you can squeeze your examples into memory then yes otherwise you need to do the splitting / merging yourself | 13:44 |
CIA-87 | shogun: Soeren Sonnenburg master * r29575c9 / examples/undocumented/csharp_modular/check.sh : | 13:49 |
CIA-87 | shogun: add proper check.sh for csharp | 13:49 |
CIA-87 | shogun: syntax is | 13:49 |
CIA-87 | shogun: ./check.sh minimal_test.cs | 13:49 |
CIA-87 | shogun: or just | 13:49 |
CIA-87 | shogun: ./check.sh - https://github.com/shogun-toolbox/shogun/commit/29575c97d1af503136bd48221cff86355e1f7a30 | 13:49 |
heiko | sonney2k, just read through KMeans::train. the method sets the cluster centers as rhs anyways | 13:49 |
@sonney2k | [1]warpy, btw I just tried - you don't need any org - whatever in csharp | 13:49 |
@sonney2k | just have a look at the minimal_test.cs example | 13:49 |
heiko | but I dont understand why rhs is used for this | 13:49 |
heiko | and not lhs | 13:50 |
@sonney2k | heiko, isn't that symmetric? | 13:50 |
heiko | oh, that might be | 13:50 |
@sonney2k | I mean after training it makes sense to change it but before - doesn't really matter | 13:50 |
heiko | what do you mean by this? | 13:51 |
heiko | oh, mmh I think this is because some classification is done | 13:52 |
heiko | something random is going on there | 13:52 |
heiko | is this to compute variances? | 13:52 |
heiko | then rhs would make sense | 13:52 |
@sonney2k | [1]warpy, and please get Load.cs to work | 13:53 |
@sonney2k | I am closing your pull request for now | 13:53 |
@sonney2k | Load.cs should just read an array | 13:53 |
@sonney2k | err read an ascii file | 13:53 |
@sonney2k | and then return arrays with numbers | 13:53 |
[1]warpy | sonney2k, what do you mean org ? | 13:56 |
@sonney2k | all the using stuff | 13:56 |
@sonney2k | all the org.shogun / org.jblas | 13:56 |
@sonney2k | laodlibrary stuff | 13:56 |
@sonney2k | should all go | 13:56 |
[1]warpy | did you pull it ? | 13:57 |
@sonney2k | look at the minimal example | 13:57 |
@sonney2k | I closed the request | 13:57 |
@sonney2k | please send updated examples | 13:57 |
[1]warpy | okay | 13:57 |
[1]warpy | i will remove them all | 13:57 |
[1]warpy | what about System.loadLibrary("modshogun"); | 14:07 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 258 seconds] | 14:25 | |
[1]warpy | sonney2k, | 14:26 |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 14:45 | |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 276 seconds] | 14:51 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 15:02 | |
@sonney2k | [1]warpy, as I said above - not needed | 15:20 |
CIA-87 | shogun: Evgeniy Andreev master * raf57ba7 / (2 files): Arrays is replaced by SGVector in CGaussianNaiveBayes - https://github.com/shogun-toolbox/shogun/commit/af57ba7c3682fbbef3c80668bee31a7f3a583ce7 | 15:21 |
CIA-87 | shogun: Soeren Sonnenburg master * rccd86e9 / (2 files): | 15:21 |
CIA-87 | shogun: Merge pull request #248 from gsomix/master | 15:21 |
CIA-87 | shogun: Arrays are replaced by SGVector in CGaussianNaiveBayes - https://github.com/shogun-toolbox/shogun/commit/ccd86e93c704b6d1e53637039ab81310cdcaa84d | 15:21 |
-!- [1]warpy [~warpy@bzq-79-181-25-215.red.bezeqint.net] has quit [Quit: HydraIRC -> http://www.hydrairc.com <- Po-ta-to, boil em, mash em, stick em in a stew.] | 16:17 | |
-!- heiko [~heiko@134.91.54.89] has left #shogun [] | 16:48 | |
-!- f-x [~user@117.192.206.212] has joined #shogun | 16:55 | |
-!- Ziyuan [~Ziyuan@116.21.254.24] has joined #shogun | 17:24 | |
Ziyuan | I opened mathematics/Math.h and got a few errors, like sqrtf cannot be resolved | 17:47 |
Ziyuan | My "math.h"'s problem? | 17:48 |
@sonney2k | Ziyuan, you need POSIX.1-2001. system | 18:57 |
-!- genix [~gsomix@95.67.178.232] has joined #shogun | 18:57 | |
-!- gsomix [~gsomix@109.169.238.89] has quit [Ping timeout: 260 seconds] | 19:00 | |
-!- in3xes_ [~in3xes@180.149.49.227] has joined #shogun | 19:00 | |
CIA-87 | shogun: Heiko Strathmann master * r63fada6 / (2 files): | 19:02 |
CIA-87 | shogun: added init method | 19:02 |
CIA-87 | shogun: made store_model_features do nothing - https://github.com/shogun-toolbox/shogun/commit/63fada6e196ece6f418e146d6b7ebd97c22783ef | 19:02 |
CIA-87 | shogun: Heiko Strathmann master * r660b180 / src/shogun/machine/DistanceMachine.cpp : store_model_features is now in header file - https://github.com/shogun-toolbox/shogun/commit/660b1805f3ba1476064b5390981dd327b952c9f9 | 19:02 |
CIA-87 | shogun: Heiko Strathmann master * rc94f370 / src/shogun/machine/DistanceMachine.cpp : added parameter include - https://github.com/shogun-toolbox/shogun/commit/c94f3705178f992d98b2f75e527933daa25d6035 | 19:02 |
CIA-87 | shogun: Soeren Sonnenburg master * r91831a1 / (2 files): | 19:02 |
CIA-87 | shogun: Merge pull request #249 from karlnapf/master | 19:02 |
CIA-87 | shogun: fix small issues - https://github.com/shogun-toolbox/shogun/commit/91831a1d5307c860f27910bc76127fe0c2e83b6b | 19:02 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Ping timeout: 240 seconds] | 19:04 | |
-!- in3xes_ is now known as in3xes | 19:07 | |
-!- genix [~gsomix@95.67.178.232] has quit [Quit: Ухожу я от вас (xchat 2.4.5 или старше)] | 19:07 | |
-!- f-x [~user@117.192.206.212] has quit [Remote host closed the connection] | 19:32 | |
-!- blackburn [~blackburn@109.226.104.206] has quit [Quit: Leaving.] | 19:58 | |
-!- blackburn [~blackburn@109.226.104.206] has joined #shogun | 20:04 | |
-!- srikanth [~mrsrikant@59.92.80.200] has joined #shogun | 20:39 | |
@sonney2k | blackburn, when you do the java tests, could you please compile with trace mallocs and then check the output at the end? | 21:10 |
@sonney2k | there should be no leaks... | 21:10 |
blackburn | yes, okay | 21:10 |
blackburn | sonney2k: have you any machine with ACML? | 21:10 |
@sonney2k | acml? | 21:11 |
blackburn | yeah, amd core math library | 21:11 |
blackburn | we have various ifdefs in lapack about ACML | 21:11 |
blackburn | sonney2k: I have stupidity assault, what is the clapack_* things in lapack.{h,cpp}? | 21:18 |
blackburn | it is instead of ATLAS, right? | 21:19 |
@sonney2k | blackburn, when atlas is not available we can still fall back to blas/lapack | 21:22 |
@sonney2k | and acml/mkl etc all are compatible (with some variance in __ naming) | 21:23 |
blackburn | aha I see | 21:23 |
-!- srikanth [~mrsrikant@59.92.80.200] has quit [Quit: Leaving] | 21:25 | |
-!- Ziyuan [~Ziyuan@116.21.254.24] has quit [Ping timeout: 258 seconds] | 21:34 | |
@sonney2k | blackburn, uh oh leaks everywhere | 21:48 |
blackburn | sonney2k: where? | 21:48 |
@sonney2k | in the modular interfaces it seems ... I tried lua and python | 21:48 |
blackburn | bad | 21:49 |
@sonney2k | the good news is that we can trace the things down easily now | 21:50 |
@sonney2k | blackburn, some of them are false alarms though... | 21:58 |
blackburn | wow I got 2x speedup for arpack haha | 22:03 |
@sonney2k | congrats! | 22:08 |
blackburn | some inverse -> LU system solving transition | 22:09 |
blackburn | haha 4x for LTSA | 22:13 |
blackburn | sonney2k: LTSA took 34s for 5000 examples | 22:15 |
-!- in3xes [~in3xes@180.149.49.227] has quit [Quit: Leaving] | 22:55 | |
-!- in3xes [~in3xes@180.149.49.227] has joined #shogun | 22:57 | |
blackburn | sonney2k: is there any way to disable atlas? :) | 23:15 |
blackburn | * removed atlas | 23:18 |
blackburn | :D | 23:18 |
@sonney2k | disable-lapack | 23:20 |
blackburn | not, need to disable atlas | 23:21 |
CIA-87 | shogun: Sergey Lisitsyn master * r2c7f253 / (2 files): Added DPOTRS and DGETRS wrappers for LAPACK - https://github.com/shogun-toolbox/shogun/commit/2c7f2531f27cc259ae43468d9f12cc34c1990d24 | 23:43 |
CIA-87 | shogun: Sergey Lisitsyn master * rf1167d6 / src/shogun/mathematics/arpack.cpp : Improved ARPACK performance by getting rid of matrix inverse - https://github.com/shogun-toolbox/shogun/commit/f1167d6343e7e947da8c92fd71deee3b3d32cbd8 | 23:43 |
--- Log closed Sat Jul 30 00:00:52 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!