--- Log opened Mon Aug 08 00:00:30 2011 | ||
CIA-87 | shogun: Soeren Sonnenburg master * r85a0cd9 / (12 files in 2 dirs): | 00:10 |
---|---|---|
CIA-87 | shogun: Merge pull request #268 from frx/streaming_vw | 00:10 |
CIA-87 | shogun: Restructured vw class definitions (+15 more commits...) - http://bit.ly/oc9DeU | 00:10 |
blackburn | sonney2k: can you please run examples/undocumented/python_modular/graphical/dimreduct/swissroll.py? | 00:27 |
blackburn | and take a screenshot or so | 00:27 |
CIA-87 | shogun: Sergey Lisitsyn master * r3f165e8 / (12 files in 2 dirs): Merge branch 'master' of github.com:shogun-toolbox/shogun - http://bit.ly/qs7cn3 | 01:18 |
CIA-87 | shogun: Sergey Lisitsyn master * r2ab04b1 / src/shogun/preprocessor/LocallyLinearEmbedding.cpp : Fixed wrong attempt to reuse matrix when LAPACK is used in LLE null space finding - http://bit.ly/rdfegh | 01:18 |
-!- f-x [~user@117.192.201.23] has quit [Remote host closed the connection] | 03:16 | |
blackburn | uh.. got it | 04:21 |
CIA-87 | shogun: Sergey Lisitsyn master * rb44191d / src/shogun/mathematics/arpack.cpp : Made ARPACK residual vector non-random - http://bit.ly/qEd5eK | 04:25 |
CIA-87 | shogun: Sergey Lisitsyn master * r015de5a / src/shogun/preprocessor/LocallyLinearEmbedding.cpp : Fixed runtime error with LLE - http://bit.ly/nngNl6 | 04:25 |
blackburn | I thought I'll never find it | 04:25 |
-!- blackburn [~blackburn@188.122.237.5] has quit [Quit: Leaving.] | 08:51 | |
-!- f-x [~user@117.192.215.3] has joined #shogun | 11:44 | |
-!- f-x [~user@117.192.215.3] has quit [Read error: Connection reset by peer] | 13:05 | |
-!- heiko [~heiko@134.91.52.156] has joined #shogun | 13:10 | |
CIA-87 | shogun: Soeren Sonnenburg master * r8730565 / (3 files in 2 dirs): | 13:14 |
CIA-87 | shogun: Merge pull request #271 from sploving/master | 13:14 |
CIA-87 | shogun: support csharp SGVector and SGMatrix and add two examples: VectorTest.cs and MatrixTest.cs - http://bit.ly/reXZYZ | 13:14 |
CIA-87 | shogun: Soeren Sonnenburg master * r8e9019a / (11 files in 3 dirs): | 13:19 |
CIA-87 | shogun: Merge pull request #270 from frx/streaming_vw | 13:19 |
CIA-87 | shogun: VW classes are now mostly SGObjects - http://bit.ly/onUf7L | 13:19 |
CIA-87 | shogun: Soeren Sonnenburg master * r96ef739 / src/Makefile : fix make clean/distclean targets - http://bit.ly/qAMCiq | 13:20 |
-!- f-x [~user@117.192.215.3] has joined #shogun | 13:27 | |
-!- blackburn [~blackburn@188.122.237.5] has joined #shogun | 14:07 | |
-!- heiko [~heiko@134.91.52.156] has quit [Ping timeout: 258 seconds] | 15:25 | |
-!- heiko [~heiko@134.91.52.156] has joined #shogun | 15:48 | |
-!- heiko [~heiko@134.91.52.156] has quit [Ping timeout: 258 seconds] | 17:02 | |
-!- blackburn [~blackburn@188.122.237.5] has quit [Quit: Leaving.] | 17:56 | |
-!- heiko [~heiko@134.91.52.156] has joined #shogun | 18:24 | |
-!- heiko [~heiko@134.91.52.156] has quit [Quit: Leaving.] | 19:21 | |
-!- heiko [~heiko@134.91.52.156] has joined #shogun | 19:21 | |
-!- heiko [~heiko@134.91.52.156] has left #shogun [] | 19:25 | |
-!- blackburn [~blackburn@188.122.237.5] has joined #shogun | 20:06 | |
-!- blackburn [~blackburn@188.122.237.5] has quit [Quit: Leaving.] | 20:18 | |
-!- blackburn [~blackburn@188.122.237.5] has joined #shogun | 20:19 | |
@sonney2k | blackburn, python swissroll.py | 21:30 |
@sonney2k | File "/usr/lib/pymodules/python2.6/matplotlib/projections/__init__.py", line 61, in get_projection_class | 21:30 |
@sonney2k | raise ValueError("Unknown projection '%s'" % projection) | 21:30 |
@sonney2k | ValueError: Unknown projection '3d' | 21:30 |
blackburn | uh | 21:30 |
@sonney2k | I guess my matplotlib is too old | 21:30 |
blackburn | old matplotlib | 21:30 |
blackburn | yes | 21:30 |
blackburn | anyway it is solved now | 21:30 |
blackburn | I've been strugelling with one bug whole night long | 21:31 |
@sonney2k | great :) | 21:32 |
@sonney2k | bug hunts can be very addictive | 21:32 |
blackburn | the problem was that arpack inits residual vector with random data | 21:32 |
blackburn | i.e. if we use one preprocessor - results are identical | 21:32 |
blackburn | but >2 preprocessors in one python script - results are sligtly different | 21:33 |
blackburn | I've checked everything | 21:33 |
blackburn | sonney2k: what to do with that script? | 21:36 |
blackburn | should I provide any action for old matplotlib? | 21:36 |
@sonney2k | blackburn, we should have 1 script per preprocessor anyways | 21:37 |
blackburn | why/ | 21:37 |
blackburn | ? | 21:37 |
@sonney2k | (like we have for kernels) | 21:37 |
blackburn | it is not test scripts | 21:37 |
@sonney2k | otherwise testing doesn't work / and users usually look for particular examples | 21:37 |
@sonney2k | we should assume newer matplotlib being available | 21:38 |
@sonney2k | which script do you refer to? | 21:38 |
blackburn | it just shows differences between preprocessors on swiss roll | 21:38 |
blackburn | what do you mean? | 21:38 |
blackburn | oh I should do KernelPCA :( | 21:39 |
@sonney2k | blackburn, name of the script? | 21:40 |
blackburn | I was talking about graphical/dimreduct/swissroll.py | 21:40 |
blackburn | and hemisphere.py too | 21:40 |
@sonney2k | ahh ok, but then it is just OK - just document the behaviour othewise others will wonder too | 21:41 |
blackburn | okay | 21:41 |
@sonney2k | blackburn, I am getting some errors compiling csharp examples now that relate to preprocessors: | 21:43 |
@sonney2k | ../../../src/interfaces/csharp_modular/BaseIsomap.cs(16,77): error CS0117: `modshogunPINVOKE' does not contain a definition for `BaseIsomap_SWIGUpcast' | 21:43 |
@sonney2k | ../../../src/interfaces/csharp_modular/modshogunPINVOKE.cs(13,7): (Location of the symbol related to previous error) | 21:43 |
@sonney2k | ../../../src/interfaces/csharp_modular/BaseIsomap.cs(16,55): error CS1502: The best overloaded method match for `DimensionReductionPreprocessor.DimensionReductionPreprocessor(System.IntPtr, bool)' has some invalid arguments | 21:43 |
@sonney2k | ../../../src/interfaces/csharp_modular/DimensionReductionPreprocessor.cs(16,12): (Location of the symbol related to previous error) | 21:43 |
@sonney2k | ../../../src/interfaces/csharp_modular/BaseIsomap.cs(16,55): error CS1503: Argument `#1' cannot convert `object' expression to type `System.IntPtr' | 21:43 |
blackburn | uh | 21:44 |
blackburn | hmm it shouldn't be BaseIsomap | 21:44 |
@sonney2k | should I do git clean dfx? | 21:45 |
blackburn | I guess it won't help | 21:45 |
blackburn | sonney2k: can I *not* describe abstract classes in *.i? | 21:46 |
blackburn | in particular, DimensionReductionPreprocessor and Isomap | 21:46 |
@sonney2k | blackburn, DotFeatures is also abstract class and in .i file | 21:46 |
@sonney2k | blackburn, did you use any templates? | 21:46 |
blackburn | DimensionReductionPreprocessor derives from SimplePreprocessor<float64_t> | 21:47 |
blackburn | stop, Isomap is not BaseIsomap | 21:47 |
blackburn | *anymore | 21:47 |
blackburn | sonney2k: so yes, try a git clean | 21:47 |
@sonney2k | :) | 21:47 |
blackburn | it seems I've changed it's name | 21:47 |
blackburn | yesterday I fixed a bug related to DSYEVR | 21:49 |
blackburn | it was a surprise I can't change matrix when calling it | 21:49 |
blackburn | oops! | 21:54 |
@sonney2k | what does that mean? | 21:54 |
blackburn | oops? | 21:58 |
blackburn | I forgot to fix similar bug in MDS :) | 21:59 |
blackburn | changing only 3 elements of LLE weight matrix leads to completely different results | 22:06 |
blackburn | for sparsity reasons.. | 22:06 |
blackburn | sonney2k: | 22:07 |
blackburn | /home/blackburn/shogun/shogun/src/shogun/classifier/vw/vw_regressor.cpp:42: undefined reference to `sg_malloc(unsigned int, char const*, int)' | 22:08 |
blackburn | /home/blackburn/shogun/shogun/src/shogun/classifier/vw/vw_regressor.cpp:46: undefined reference to `sg_calloc(unsigned int, unsigned int, char const*, int)' | 22:08 |
f-x | looks like it's in my code | 22:08 |
@sonney2k | blackburn, how did you get that? | 22:08 |
@sonney2k | I just compiled from scratch and I don't have this | 22:08 |
blackburn | f-x: yeah sorry didn't noticed you are here ;) | 22:09 |
blackburn | what is the options you have used? | 22:09 |
@sonney2k | ./configure --interfaces=csharp_modular,python_modular --enable-debug --prefix=/usr/local/stow/shogun-0.99.9 --disable-optimization | 22:09 |
blackburn | I tried without --enable-debug and got it | 22:09 |
blackburn | will try now with | 22:09 |
@sonney2k | blackburn, do git clean dfx | 22:10 |
@sonney2k | the file got renamed... | 22:10 |
@sonney2k | (no longer exists) | 22:11 |
blackburn | ah | 22:11 |
blackburn | okay | 22:11 |
CIA-87 | shogun: Sergey Lisitsyn master * r63e6105 / src/shogun/preprocessor/ClassicMDS.cpp : Fixed wrong attempt to reuse matrix while DSYEVR is called - http://bit.ly/nIwrp6 | 22:21 |
@sonney2k | f-x, with size_t transition you mean what exactly? | 22:37 |
@sonney2k | use index_t ? | 22:38 |
f-x | sonney2k: what should size_t become? | 22:38 |
f-x | index_t is valid in most cases, true | 22:38 |
@sonney2k | it depends - most time index_t | 22:38 |
f-x | but one or two places the code expects size_t to be unsigned | 22:38 |
f-x | i think that is quite minor - i can replace it with uint32_t in those places | 22:38 |
@sonney2k | f-x, or keep size_t if it is using official functions like memcpy or so | 22:49 |
f-x | sonney2k: there are many custom functions returning size_t too | 22:51 |
f-x | in vw | 22:51 |
f-x | i'll have a look into it soon.. uint32_t appears safe, i'll use index_t where more appropriate | 22:52 |
f-x | sonney2k: if it put it in lib/Hash.*, do I have to make a class for this? | 22:53 |
@sonney2k | f-x, no | 22:54 |
@sonney2k | there is lib/Hash.* already | 22:54 |
f-x | ah | 22:54 |
f-x | ok | 22:54 |
@sonney2k | just add the needed functions | 22:54 |
@sonney2k | with proper names (just look at the code it is obvois) | 22:54 |
f-x | sonney2k: ok | 22:55 |
--- Log closed Tue Aug 09 00:00:42 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!