Open in new window / Try shogun cloud
--- Log opened Mon Jan 30 00:00:19 2012
-!- blackburn [~qdrgsm@] has quit [Quit: Leaving.]00:28
-!- blackburn [~qdrgsm@] has joined #shogun03:47
CIA-11shogun: Sergey Lisitsyn master * r689b5ed / (6 files): Fixes and improvements of EDRT -
shogun-buildbotbuild #465 of octave_static is complete: Success [build successful]  Build details are at
CIA-11shogun: Sergey Lisitsyn master * ra253451 / (6 files in 2 dirs): CoverTree for EDRT -
-!- blackburn [~qdrgsm@] has quit [Quit: Leaving.]07:27
-!- blackburn [~qdrgsm@] has joined #shogun07:59
CIA-11shogun: Soeren Sonnenburg master * r3da3d15 / src/shogun/ui/GUIConverter.cpp : fix compilation error when lapack is not available -
CIA-11shogun: Soeren Sonnenburg master * race4ba9 / (3 files in 3 dirs):10:54
CIA-11shogun: include <clapack.h> directly10:54
CIA-11shogun: we cannot include <atlas/clapack.h> as some distriutions ship clapack10:54
CIA-11shogun: under /usr/include -
CIA-11shogun: Soeren Sonnenburg master * r66ebf04 / src/shogun/features/BinnedDotFeatures.cpp : fix normalization in binned dot features' dot() function -
CIA-11shogun: Soeren Sonnenburg master * r7827b83 / (2 files):10:55
CIA-11shogun: add SGVector based dense_dot_vec function wrapping dense_dot to10:55
CIA-11shogun: dotfeatures -
-!- f-x [~user@] has joined #shogun14:16
blackburnsonne|work: around?14:31
blackburnwe've got many cache misses in distance and kernel calculation14:37
blackburnit is bad to do d[i,j] and d[j,i] in the same time14:37
blackburnmatrix calculation I mean14:38
-!- f-x [~user@] has quit [Remote host closed the connection]14:43
-!- f-x [75d89915@gateway/web/freenode/ip.] has joined #shogun14:47
-!- f-x [75d89915@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]14:58
-!- Ram108 [~amma@] has joined #shogun15:23
-!- cronor [] has joined #shogun15:49
sonne|workblackburn: agreed16:24
blackburnI wrote too much to understand what is you agree with :)16:25
sonne|workit is bad to do d[i,j] and d[j,i] in the same time16:25
blackburnah yes16:25
sonne|workdid you write more than that?16:25
blackburnwell yes, for last two days16:25
blackburnsonne|work: I can hardly wait to release16:26
blackburncause I can't post on mloss and submit paper16:27
blackburnblocks me a little16:27
sonne|workheikos stuff needs testing before we can16:27
sonne|workthen do a snapshot16:27
blackburnof current version?16:27
sonne|workIIRC make release SNAPSHOT=yes16:28
blackburnhmm makes sense16:28
blackburnsonne|work: have you seen I did integrate covertree?16:31
blackburnI know you probably not a big fan of std things but it works smooth, that's definitely better than nothing16:31
sonne|workno havent seen it16:31
sonne|workno time now16:31
blackburnsonne|work: well then you know it now :)16:32
sonne|workwhere is it?16:32
-!- Ram108 [~amma@] has quit [Ping timeout: 248 seconds]16:52
@sonney2kblackburn, what should index_t be?16:57
blackburnsonney2k: unsigned?16:57
@sonney2kwell possible if you are carefull with signedness everywhere16:58
@sonney2kwould probably be a big change16:58
blackburnI think we should -> to unsigned some day16:58
@sonney2kblackburn, do you know whether gsoc got announced already?16:58
blackburnsonney2k: not yet16:58
@sonney2kthen things like for int i=0 i< size ; ... won't work16:59
blackburnsonney2k: why?16:59
@sonney2kwe would have to go through all loops and convert them to for index_t i=0 ...16:59
blackburnhmm yes, but not a big change16:59
@sonney2ksome of these are doing plus minus operations like17:00
blackburncan be done gradually17:00
@sonney2kand check if things are >017:00
@sonney2kcan be done but lot of work17:00
blackburnsonney2k: it leaves room for optimization I think, i.e. when compiler knows i is unsigned17:00
@sonney2kwhat does it help?17:00
blackburnI'm not sure though17:01
blackburndon't you think it would be better for compiler17:01
@sonney2kunderneath (machine language) signed / unsigned is no difference17:01
blackburnwell, for example loop to zero is faster17:02
@sonney2ksame thing17:03
blackburnsonney2k: I am not sure, zero test could be faster :)17:05
blackburnsonney2k: have you seen my benchmark vs scikits?17:05
@sonney2kunsigned int u=0 ====== int u=0 :-)17:05
blackburnfinal one17:05
@sonney2kall bits 017:05
blackburnyou could check if you want17:05
blackburnin the dropbox paper17:05
blackburnin the dropbox folder17:06
blackburnnow with fruits^W error bars17:10
CIA-11shogun: Soeren Sonnenburg master * ra29c0a6 / (7 files in 4 dirs): fix various doxygen documentation warnings -
-!- Ram108 [~amma@] has joined #shogun17:20
CIA-11shogun: Sergey Lisitsyn master * r2443b79 / (2 files): Added covertree libshogun example -
blackburnsonney2k: what is O9 flag?17:21
blackburnwhy not highest O3?17:23
-!- Ram108 [~amma@] has quit [Ping timeout: 244 seconds]17:44
-!- Ram108 [~amma@] has joined #shogun18:11
-!- f-x [75c0d948@gateway/web/freenode/ip.] has joined #shogun18:27
-!- f-x [75c0d948@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:57
-!- cronor [] has quit [Ping timeout: 252 seconds]19:03
-!- blackburn [~qdrgsm@] has quit [Quit: Leaving.]20:00
-!- blackburn [~qdrgsm@] has joined #shogun20:04
-!- Ram108 [~amma@] has quit [Ping timeout: 260 seconds]20:57
shogun-buildbotbuild #134 of nightly_none is complete: Success [build successful]  Build details are at
CIA-11shogun: Soeren Sonnenburg master * r67ea574 / src/shogun/converter/Isomap.cpp : remove unused variables j and tmp from Isomap to fix warning -
CIA-11shogun: Soeren Sonnenburg master * r9df3f64 / src/shogun/features/BinnedDotFeatures.h : document remaining methods of binned dot features -
CIA-11shogun: Sergey Lisitsyn master * r27b603d / examples/undocumented/libshogun/Makefile : Added check-valgrind target for libshogun makefile -
CIA-11shogun: Sergey Lisitsyn master * r03e8d6b / examples/undocumented/libshogun/library_gc_array.cpp : Fixed GCArray libshogun example -
shogun-buildbotbuild #471 of octave_static is complete: Failure [failed test_1]  Build details are at  blamelist: sonne@debian.org22:14
shogun-buildbotbuild #472 of octave_static is complete: Success [build successful]  Build details are at
CIA-11shogun: Sergey Lisitsyn master * r03786e6 / (3 files): Proper automatic k neighborhood matrix handling for LLE,HLLE,LTSA -
CIA-11shogun: Sergey Lisitsyn master * r52334da / src/shogun/converter/KernelLocallyLinearEmbedding.cpp : Covertree max distance computation fix for KLLE -
--- Log closed Tue Jan 31 00:00:19 2012