Shogun Machine Learning - IRC Logs
--- Log opened Fri Jun 02 00:00:55 2017
-!- mikeling [uid89706@gateway/web/] has joined #shogun03:33
mikelingwiking: ping03:33
mikelingare you still there?03:34
shogun-buildbotbuild #75 of nightly_all is complete: Success [build successful]  Build details are at
-!- mikeling [uid89706@gateway/web/] has quit [Quit: Connection closed for inactivity]08:03
-!- mikeling [uid89706@gateway/web/] has joined #shogun08:11
-!- geektoni [] has joined #shogun09:06
@wikingmikeling, sorry09:14
@wikingstill CET... starting from next week i'm UTC+809:14
@wikingmikeling, i'm here now09:15
mikelingwiking: np, don't mind :)09:28
mikelingBTW, the will return a reference of that data element09:28
@wikingbut you can use operator[]09:29
@wikingi.e. vector[i] as well09:29
@wikingit's the same09:29
mikelinghmm, how could I make it return a const reference for
@wikingif you return vector[i]09:30
@wikingthat should be fine09:30
@wikingit'll transform it to be const09:30
@wikingand vector<T>[i] or vector<T>.at(i) returns T&09:30
@wikinglemme get you the link09:31
@wikingit has two different version of09:31
@wikingand if yo udo there return vector[i];09:32
@wikingwhere the return type of the vector is09:32
@wikingconst T&09:32
mikelingit's same like at()
@wikingthen the compiler will use const_reference version of the operator[]09:32
@wikingbut honestly09:32
@wikingnobody uses .at for vector09:32
@wikingor it's rather seldom that you see people use it09:32
@wikingit's rather the operator[] that is being used09:33
@sukeyNew Commit "[ProgressBar] Fix warning when building." to shogun-toolbox/shogun by vigsterkr:
micmnwiking: regarding my pr09:40
micmnhow do you think we should proceed09:41
micmnfor the regularization param09:41
micmnand the template?09:41
mikelinga second09:41
mikelingsorry, I just thought you are @me ;)09:43
mikelingwiking: Do you have any idea about why this test error out?
mikelingI'm working on it since yesterday, but found nothing for it09:49
@sukeyPull Request #3828 "[ProgressBar] Add multi threaded progress bar."  synchronized by geektoni -
-!- iglesiasg [~iglesiasg@] has joined #shogun10:06
-!- mode/#shogun [+o iglesiasg] by ChanServ10:06
-!- Raman [3bb2b171@gateway/web/freenode/ip.] has joined #shogun10:12
RamanI am a beginner and want to do some contribution.10:13
RamanCan anyone suggest how and where I can start?10:13
geektoniHi Raman :) please take a look to the project wiki -->
-!- Raman [3bb2b171@gateway/web/freenode/ip.] has quit [Quit: Page closed]10:26
-!- travis-ci [] has joined #shogun10:33
travis-ciit's Giovanni De Toni's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []10:33
@wikingmicmn, hey11:17
@wikingmicmn, i mean if you are blocked and dont wanna hack too much with the linalg then you can just put a todo to the code11:17
@wikingand continue on LDA11:17
@wikingotherwise the ideal thing would be that you implement whatever is still missing from linalg that is needed for the lda refactor11:18
@wikingmikeling, checking11:18
micmnyeah that's not a problem11:18
@wikingoh nooo i know11:18
@wikingmikeling, yeah i owe you this11:18
@wikinglemme do it asap11:18
@wikingmikeling, can i ask you a favour plz11:19
@wikinglemme do it11:19
mikelingwiking: sure11:19
mikelingwhat's up11:19
@wikingnothing i can cherrypick your repo :P11:19
micmnmy question was: how should we proceed? join the 2 classes or implement regularization and templating in FisherLDA too?11:20
@wikingi'm wondering whether we really need 2 different algos11:25
@wikingi mean classes11:25
@wikingof course its actually nicer in a way to maintain code like that11:25
@wikingthat's the only reason i would say keep 2 different classes and implement templating and reg there as well (fisherlda)11:26
@wikingbecause i really hate seeing source files more than 150+ LoC11:26
@wikingmicmn, do you have any strong pros for merging the 2 classes?11:30
micmnit depends on what we want to do we the lda's solver11:31
micmnif we want to keep it then 2 classes are fine11:31
micmnotherwise i think that fisherlda's solver need to handle the 2 classes case as a special case11:32
micmnso it breaks a bit the abstraction11:32
micmni mean for the bias computation11:33
micmnand actually i think for the weights too but i have to check11:33
@wikingmikeling, still building it locally11:43
micmnis there a nice way with SGMatrix/Vector to do matrix.col(i) -= vec?11:48
@wikingmicmn, linalg should do that11:49
@wikingbecause we wanna get out from SGMatrix all those operators11:49
@wikingand handle it with linalg11:49
-!- mikeling is now known as mikeling|brb11:49
micmn*adds to todo list*11:50
@wikingmikeling|brb, ok so this ctest is acting a bit weird11:53
@wikingbecause if i run ctest -R unit-SGObject11:54
@wikingi get 232 FAILED TESTS11:54
@wikingbut if i run one of those failed tests separately11:54
@wiking/home/wiking/shogun/build/bin/shogun-unit-test --gtest_filter=SGObject.clone_equals_StreamingStringFeatures_PT_CHAR11:55
@wikingit does not fail11:55
@wikingit only fails consequently with11:55
@wiking/home/wiking/shogun/build/bin/shogun-unit-test --gtest_filter=SGObject.equals_DynamicObjectArray_equal11:55
-!- iglesiasg [~iglesiasg@] has quit [Remote host closed the connection]11:56
@wikingmikeling|brb, i've added the comment that should start fixing stuff :)12:10
@wikingi'm just recompiling12:10
@wikinglocally to test it12:11
@wikingmikeling|brb, the commented code fixed most of the errors12:16
@wikingthere's still error in case of SGObject.equals_DynamicObjectArray_different12:16
@wikingand for that i suggested in the comments12:24
@wikinghow you  could fix it12:24
@sukeyPull Request #3828 "[ProgressBar] Add multi threaded progress bar."  merged by vigsterkr -
@sukeyNew Commit "Merge pull request #3828 from geektoni/feature/progress_multi12:39
@sukey[ProgressBar] Add multi threaded progress bar." to shogun-toolbox/shogun by vigsterkr:
@wikinggeektoni, done12:39
-!- iglesiasg [~iglesiasg@] has joined #shogun13:18
-!- mode/#shogun [+o iglesiasg] by ChanServ13:18
-!- mikeling|brb is now known as mikeling14:48
mikelingwiking: thank you! I'm checking14:49
@wikinglemme know if there's something that's not clear15:01
@wikingbut it should be rather straightforward to fix15:01
micmnwiking: besides the column+vector op i need trace, identity and zero (already in SGMatrix), should i add them to linalg?15:21
micmnA-B = linalg::add(A,B,1,-1)?15:22
@wikingi hope so :)15:23
micmnit's a bit ugly but ok :D15:24
@sukeyPull Request #3829 "[ProgressBar][WIP] Add a boolean flag to the progress bar."  opened by geektoni -
@sukeyPull Request #3829 "[ProgressBar][WIP] Add a boolean flag to the progress bar."  synchronized by geektoni -
-!- iglesiasg [~iglesiasg@] has quit [Quit: leaving]16:57
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has quit [Ping timeout: 260 seconds]17:02
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has joined #shogun17:14
@sukeyPull Request #3829 "[ProgressBar][WIP] Add a boolean flag to the progress bar."  synchronized by geektoni -
@sukeyPull Request #3830 "Add to linalg trace, zero, identity, add_col_vec (WIP)"  opened by micmn -
@sukeyPull Request #3826 "Remove duplicate code in LDA/FisherLDA (WIP)"  synchronized by micmn -
-!- geektoni [] has left #shogun []19:05
-!- mikeling [uid89706@gateway/web/] has quit [Quit: Connection closed for inactivity]20:33
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has quit [Ping timeout: 260 seconds]21:01
-!- WangWang [uid231047@gateway/web/] has quit [Quit: Connection closed for inactivity]21:09
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has joined #shogun22:29
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has quit [Quit: Page closed]23:50
--- Log closed Sat Jun 03 00:00:57 2017