Open in new window / Try shogun cloud
--- Log opened Tue Apr 09 00:00:10 2013
--- Day changed Tue Apr 09 2013
-!- lv [] has joined #shogun00:00
-!- lv [] has left #shogun []00:00
-!- medeeiip [~medeeiip@] has quit [Read error: Connection reset by peer]00:02
-!- bogdanc [] has quit [Ping timeout: 276 seconds]00:09
-!- arleyristar [b183f4d8@gateway/web/freenode/ip.] has joined #shogun00:18
-!- jptech93 [~quassel@] has joined #shogun00:27
naywhayareblackburn: you weren't kidding about the 10 billion people01:24
naywhayarethe swarm is descending upon me01:24
n4nd0naywhayare: you better design a good classifier to help you ;)01:24
-!- praveen97uma [6adb6aa5@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]01:28
n4nd0sonney2k: page updated with participation announcement01:37
n4nd0this admin interface editor gets me on my nerves :S01:37
-!- n4nd0 [] has quit [Quit: leaving]01:39
-!- medeeiip [~medeeiip@] has joined #shogun02:00
-!- debajyoti [c10b818b@gateway/web/freenode/ip.] has joined #shogun02:19
debajyotihi, I am new here. can someone guide me with subscribing to the mailing list? thanks02:27
medeeiipsend an empty mail to shogun-list-subscribe@shogun-toolbox.org02:30
debajyotitried that about 2 hours ago.02:31
debajyotii tried to post, but failed now02:31
debajyotishould i give it more time?02:31
medeeiipI was going by the code for handling files. For handling files c style file pointer is used. Is there any special reason for doing this02:32
medeeiipeven in the functions like get_matrix() reference is passed as arguments... rather I can return a pair of references to the matrix and it's dimenssion02:36
-!- foulwall [] has joined #shogun02:38
medeeiipI'm just asking about opinion....02:40
-!- medeeiip [~medeeiip@] has quit [Quit: Leaving]02:47
debajyotimy problem has been solved, i can access the mailing list now. thanks02:47
-!- phoenix_ukr [] has quit [Ping timeout: 264 seconds]02:57
-!- subramanian_v [b89439dd@gateway/web/freenode/ip.] has joined #shogun02:59
subramanian_vHello Everybody03:06
subramanian_vI looked at shogun GSoC page03:06
subramanian_vSo what kind of optimization algorithms should we know ?03:07
-!- zxtx [~zv@] has quit [Ping timeout: 256 seconds]03:18
subramanian_vanybody there03:27
-!- arleyristar [b183f4d8@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]03:36
-!- debajyoti [c10b818b@gateway/web/freenode/ip.] has quit [Quit: Page closed]03:39
-!- subramanian_v [b89439dd@gateway/web/freenode/ip.] has quit [Quit: Page closed]03:43
-!- subbu [b89439dd@gateway/web/freenode/ip.] has joined #shogun03:44
-!- foulwall [] has quit [Ping timeout: 245 seconds]03:49
subbuanybody online ?04:01
-!- ujjwal [c77e807e@gateway/web/freenode/ip.] has joined #shogun04:05
-!- maheshakya [~chatzilla@] has joined #shogun04:36
-!- subbu [b89439dd@gateway/web/freenode/ip.] has quit [Quit: Page closed]04:42
-!- subbu [b89439dd@gateway/web/freenode/ip.] has joined #shogun04:48
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun05:05
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun05:17
-!- maheshakya [~chatzilla@] has left #shogun []05:38
blackburnnaywhayare: I just read your application form - are you planning to get a lot of guys who know SFINAE and CRTP? :D05:43
naywhayareblackburn: no, I don't expect to, but how people answer those questions will tell me a lot about them (if they choose to answer) and that will help with the filtering05:50
naywhayarenote that one of the "challenge questions" is actually the halting problem :)05:50
naywhayarethe responses to that, if people respond, should be entertaining, but they should also tell us a lot about who to avoid...05:51
blackburnnaywhayare: I expect quite a few will answer you with something reasonable :)05:51
naywhayareI would have made the application less intimidating (I'm not sure intimidating is the right word), but considering that I already have a bunch of people begging for more information, some amount of filtering seems necessary05:52
-!- subbu [b89439dd@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]05:53
naywhayareanyway, it's bedtime now.  'night :)05:53
blackburnnaywhayare: morning for me - good night ;)05:53
-!- foulwall [] has joined #shogun06:50
-!- bogdanc [] has joined #shogun07:17
-!- foulwall [] has quit [Remote host closed the connection]07:27
-!- gsomix [~Miranda@] has quit [Quit: Miranda IM! Smaller, Faster, Easier.]07:44
-!- foulwall [~foulwall@2001:da8:215:6901:d4bb:90e5:e26c:36a4] has joined #shogun07:56
-!- jptech93 [~quassel@] has quit [Remote host closed the connection]08:04
-!- jptech93 [~quassel@] has joined #shogun08:05
-!- phd [] has joined #shogun08:28
-!- phoenix_ukr [] has joined #shogun08:28
-!- phoenix_ukr [] has left #shogun []08:29
-!- ChanServ changed the topic of #shogun to: "Shogun Machine Learning Toolbox | Devs are usually around between 10am-10pm UTC | Workshop upcoming July 12-14, Berlin C-Base | We participate in GSoC 2013 | When asking questions please be patient or use the mailing list. We always answer but it might take a while. | Channel logs"08:34
@sonney2kblackburn, are you awake?08:39
@sonney2kwait you have been awake whole night?!08:39
-!- hoijui [] has joined #shogun08:46
-!- n4nd0 [] has joined #shogun08:46
n4nd0sonne|work: good morning!08:47
n4nd0I had an idea, we could add an initiation task in github suited for the people interested in the demos/django project08:49
n4nd0to integrate the history revision in the admin interface :)08:49
-!- akshay_hebbar [~akshay@] has joined #shogun08:50
n4nd0providing that it seems several people are interested in this project, maybe someone can get it done soon. I think we sort of *need* this feature08:50
n4nd0sonne|work: if you think it is fine I will write a github issue08:50
-!- zxtx [] has joined #shogun08:57
blackburnsonney2k: was accidentally awake early in the morning then got asleep again :)08:59
sonne|workn4nd0: hey there!08:59
sonne|workblackburn: I was already wondering :D08:59
blackburnn4nd0: I think that's fine09:00
sonne|workn4nd0: IDK09:00
blackburnhowever no idea how to organize that09:01
-!- hoijui [] has quit [Read error: Operation timed out]09:02
n4nd0sonne|work: why not?09:02
-!- foulwall [~foulwall@2001:da8:215:6901:d4bb:90e5:e26c:36a4] has quit [Remote host closed the connection]09:05
sonne|workn4nd0: because it is totally unrelated to the task - then I would rather want them to directly do stuff in d3js or some form via which one can upload data to the shogun server09:06
n4nd0sonne|work: mmm I see. I though django was also important for this project.09:07
-!- phd [] has quit [Ping timeout: 256 seconds]09:08
sonne|workn4nd0: yes that is in django09:12
sonne|workbut working one some stuff on the website - hmmhh totally unrelated to ML and core shogun isn't such a nice thing09:12
sonne|workIMHO this just has to be done and people should rather do ML & real applications09:12
n4nd0sonne|work: Ok, I understand your point09:13
blackburn I should be careful in that city :D09:14
sonne|workis she ok?09:15
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has joined #shogun09:16
blackburnsonne|work: no idea, probably09:16
-!- medeeiip [~medeeiip@] has joined #shogun09:20
blackburnsonne|work: do you know aanything on SOCP?09:21
medeeiipI was going by the code for handling files. For handling files c style file pointer is used. Is there any special reason for doing this09:21
blackburnmedeeiip: c style file pointer? what do you mean?09:22
sonne|workblackburn: yes sure09:22
medeeiipi mean FILE* f; type handling of any file09:22
medeeiipno use of fstream09:23
blackburnsonne|work: because I have nothing to answer to cheng :)09:23
blackburnmedeeiip: I see09:23
blackburnmedeeiip: I don't think there is a special reason to do that09:24
medeeiip even in the functions like get_matrix() reference is passed as arguments through which value is returned... rather I can return a pair of references to the matrix and it's dimenssion09:25
-!- jptech93 [~quassel@] has quit [Ping timeout: 245 seconds]09:25
-!- gzhd79 [] has joined #shogun09:26
-!- jptech93 [~quassel@] has joined #shogun09:26
blackburnmedeeiip: why to create a new object to handle that?09:26
sonne|workmedeeiip: we had typemaps so we could use python files directly in shogun09:27
sonne|workmedeeiip: mapping FILE* <-> python09:27
medeeiipof course there will be change in ovarall code.. But using fstream implementing program(program using shogun library) will be a lot safer..... vary low chance of memory leak09:29
medeeiip@blackburn you mean fstream?09:30
blackburnmedeeiip: I mean pair09:30
medeeiipbecause pair is lightweight and can return multiple things at once.....09:31
medeeiipi feel in todays computer this much memory would put a constrient09:32
blackburnwhat is that lightweight in pair? it is an object09:32
medeeiipi mean wouldn't@blackburn09:32
medeeiipan object containing only two reference09:32
blackburnpair do not contain references09:33
-!- jptech93 [~quassel@] has quit [Ping timeout: 264 seconds]09:33
blackburnit contains objects09:33
blackburnor values of PODs09:33
medeeiipif i return a pair like pair<int&,int&> what is wrong?09:34
blackburnthat doesn't sound much safer than what we have09:34
-!- foulwall [] has joined #shogun09:35
medeeiipi'm not saying pair is safer..........09:35
medeeiipI'm saying using fstream instade of FILE is safer09:35
medeeiipand pair is simple ...and easy to work with09:36
blackburnmedeeiip: so I have to rewrite it as you don't like it, right?09:36
medeeiipI'm not saying that..... I'm just taking opinion.... That's all09:37
medeeiipshogun has posted a idea for gsoc about improving file support.... For this what specifically you are asking for?09:38
medeeiipThe post said overall improvement of code quality09:38
blackburnyes and if you have suggestions feel free to make them come true :)09:39
medeeiipthat's ok... but what your org mean to say by saying improvement of overall code quality09:40
blackburnmedeeiip: changing it to fstream might be an example of improving code quality09:41
medeeiipthank you 'n btw one more general question: where buglist of shogun is mentained?09:43
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]09:43
blackburnmedeeiip: github issues09:43
blackburnmedeeiip: the problem with legacy things like FILE is that we don't really have time to rewrite code09:44
blackburnand if we rewrite it we should cover it with unit-tests09:44
medeeiipyou mean file handling09:45
blackburnmedeeiip: that's not the only example of not-the-best-C++-practice09:45
blackburnbut we are more or less sure it is working so changing working code with prettier code is not the way to go - it should be properly tested then09:46
medeeiipI'mean you are asking unit tests for what?09:46
blackburnmedeeiip: I mean if we change code we need unit-tests09:46
medeeiipGot it......09:47
blackburnso far we had no power neither to rewrite code nor write unit-tests for that09:47
blackburnthat's why we hope students will help to improve the code09:47
blackburnhow what?09:48
medeeiiphow you had no power to rewrite code?09:48
blackburnthere are a few of us only here and we have real jobs :)09:49
blackburnbe back soon09:49
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]09:49
medeeiipsee you09:49
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has joined #shogun09:54
-!- medeeiip [~medeeiip@] has quit [Ping timeout: 256 seconds]09:57
-!- bogdanc [] has quit [Ping timeout: 256 seconds]10:06
-!- medeeiip [~medeeiip@] has joined #shogun10:08
-!- ppletscher [] has joined #shogun10:08
-!- medeeiip [~medeeiip@] has quit [Client Quit]10:08
sonne|workn4nd0: look pluskid wrote some nice summary for his last years project
sonne|workI cannot parse it though :D10:09
n4nd0sonne|work: hehe yeah, me neither10:09
n4nd0hi ppletscher! nice to see you around here :)10:09
n4nd0I am guessing you are Patrick, maybe I got it wrong though hehe10:10
ppletscherhi all, thought I give this irc a shot. Yes, I'm Patrick10:10
-!- blackburn [] has joined #shogun10:10
n4nd0ppletscher: have many students contacted you already?10:10
blackburnppletscher: nice to see you here!10:11
ppletscherno, I didn't get any emails yet10:11
blackburnppletscher: I am sorry to say but you may receive quite a few so be ready ;)10:11
ppletscheralright, looking forward to the applications10:11
n4nd0ppletscher: aham! Ok, I had planned to do some preliminary work in the project to support my application so I will probably write to you soon10:12
ppletschern4nd0: sure, just shoot me an email10:13
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]10:14
-!- rishabh [~rishabh@] has joined #shogun10:17
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit []10:19
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has joined #shogun10:20
sonne|workppletscher: lucky you ;)10:20
ahcordeHello, I'm Alex, I'm studying a Master of Computer Vision. Few months ago I forked the repository on github because I used this library to develop some of my pattern recognition practices. I would like to start to help in this library. I visit in github project page the issues and I'm try to patch the problem with json serialization. I don't know if someone can tell me which is the best way to start to develop in Shogun.10:21
blackburnahcorde: the best way is to find something you can fix/improve10:21
blackburnand do that :)10:21
n4nd0ahcorde is also one of my dearest friends from uni :)10:22
blackburnI am sure we have a great room for improvement so it should not be a problem10:22
blackburnn4nd0: nice10:23
-!- bogdanc [~bogdan@] has joined #shogun10:23
n4nd0ahcorde: have you taken a look to sth related to this json serialization problem? The github issue is not very descriptive indeed :D10:24
-!- bogdan__ [~bogdan@] has joined #shogun10:40
-!- bogdanc [~bogdan@] has quit [Ping timeout: 245 seconds]10:41
-!- foulwall [] has quit [Ping timeout: 276 seconds]10:48
-!- bogdan__ [~bogdan@] has quit [Ping timeout: 248 seconds]11:02
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:05
ahcordeI have tried to debug json serialization example11:08
sonne|workahcorde: and?11:08
-!- gaurang [] has joined #shogun11:08
sonne|workahcorde: it is totally unclear why it stopped ...11:08
sonne|workahcorde: in worst case you have to go back to shogun 1.0 and do a git bisect to figure out where it broke11:12
-!- bogdan__ [~bogdan@] has joined #shogun11:12
-!- ahcorde [c1934d18@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]11:18
-!- bogdan__ [~bogdan@] has quit [Ping timeout: 252 seconds]11:20
-!- bogdan__ [~bogdan@] has joined #shogun11:21
-!- bogdanc [~bogdan@] has joined #shogun11:24
-!- bogdan__ [~bogdan@] has quit [Ping timeout: 245 seconds]11:27
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun11:28
-!- bogdan__ [~bogdan@] has joined #shogun11:29
-!- ahcorde [c1934745@gateway/web/freenode/ip.] has joined #shogun11:31
-!- bogdanc [~bogdan@] has quit [Ping timeout: 276 seconds]11:33
deerishii would like to contibute to Blind Source Separation (BSS) and Independent Component Analysis (ICA) in the shogun ml toolkit.i know Machine Learning,Natural language processing and Signal do i start?any suggestions?11:35
-!- heiko [] has joined #shogun11:35
blackburnheiko: heiko heiko11:36
heikohi blackburn :)11:37
heikohow is it going?11:37
blackburnheiko: we didn't find the genius you need yet ;)11:37
heikoblackburn: what do you  mean?11:37
blackburndeerishi: as usual - start contributing and learning more on the task11:37
heikojust read the happy news! :)11:37
blackburnheiko: I am still joking about your idea11:37
blackburnheiko: well if you didn't read the mail you could notice the # of people here11:38
-!- bogdanc [~bogdan@] has joined #shogun11:40
heikoblackburn: yes I noticed11:43
heikothey are all up for the visualisation task11:43
heikothis is good!11:43
heikoblackburn: well we have lambday right?11:43
blackburnheiko: haha11:43
heikoI haven't been here recently but he seemed quite comitted11:44
blackburnheiko: oh well we have 1.5 months or so11:44
heikoand actually the first person who solved an intro task11:44
-!- bogdan__ [~bogdan@] has quit [Ping timeout: 252 seconds]11:44
heikoblackburn: yes, so thats fine11:44
heikoblackburn, sonne|work we should update the chat log for gsoc11:44
heikoand maybe put a little link on the mainpage of the website11:45
blackburnheiko: how?11:45
blackburnheiko: what is wrong?11:45
heikomention gsoc11:45
heikoits only on the events page11:45
n4nd0heiko: it is on the main page as well11:45
heikon4nd0: not here11:46
n4nd0ah BTW, I wrote yesterday a couple of tweets in ShogunToolbox, so please retweet them11:46
heikon4nd0: I dont use twitter apart from shogun :)11:46
sonne|workheiko: what?11:46
heikobut thanks for doing it11:46
heikoso I dont see gsoc on the mainpage, I have to click on news11:46
n4nd0heiko: check in the middle
heiko :)11:46
n4nd0heiko: :)11:47
sonne|workheiko: but you did see the workshop did you?11:47
n4nd0sonne|work: I wrote in my blog about gsoc too11:47
heikosonne|work: yes, thats what distracted me :)11:47
sonne|workn4nd0: very good!11:47
n4nd0sonne|work: in case you want to reblog or sth11:47
sonne|workheiko: IDK how many students did contact you but I am swamped11:47
n4nd0sonne|work: haha all of them for demos?11:48
blackburnsonne|work: demoooos more demoos11:48
heikosonne|work, not a single one so far :)11:48
heikowe can have two demo projects11:48
n4nd0or 811:48
heikoone for this internet stuff11:48
sonne|worklarge scale11:48
sonne|workdemos, file IO11:48
heikoand one only for cool visualisations11:48
blackburnI didn't receive any email on dim reduction11:48
sonne|workmaybe due to the natural up -> down reading11:48
n4nd0blackburn: I will mail you to cheer you up if you want :P11:49
n4nd0haha that would be pretty lame, if we actually just get applications for the proposals on top because people don't read all11:49
blackburnn4nd0: no we end up with a lot of applications11:49
heikoblackburn, sonne|work just received the first mail :_)11:55
heikoits from lambday :)11:55
blackburnheiko: I am totally alone *crying*11:55
-!- bogdanc [~bogdan@] has quit [Ping timeout: 264 seconds]11:55
blackburnheiko: okay actually I received CCs for andreas' and  cheng's ideas11:56
heikoblackburn:  you will be fine!11:57
sonne|workblackburn: so you have more time doing the admin job!11:57
blackburnsonne|work: heiko: BSS is quite popular11:57
heikoblackburn: btw last week I heard a talk on kernel ECA11:57
heikodo you know that?11:57
heikosounded cool11:57
blackburnheiko: no, what is it?11:57
-!- gaurang [] has quit [Remote host closed the connection]11:57
sonne|workblackburn: I know...11:57
blackburnentropy component analysis?11:57
heikoEntropy component analysis11:57
sonne|workheiko: kernel ECA?11:57
heikoinstead of taking the eigenvectors with most signal in the RKHS11:57
heikoyou take those that maximise the entropy in the *input* space11:58
heikowith this you get quite different results11:58
heikosonne|work: yes11:58
blackburnheiko: do you have any good reference?11:58
sonne|workheiko: hmmhh so it finds you the most relevant input dims wrt the kernel used?11:59
heikosonne|work: kernel acts as a smoothing operator on the distribution in the input space, Ill search a reference11:59
heikothey define some entropy using the kernel11:59
heikoso there is the connection12:00
blackburnheiko: I am a bit lost12:01
blackburnthey speak about RKHS and parzen window12:01
heikoblackburn:  there might be more papers on this on Robert Jenssens website12:02
heikothis is the first I guess12:03
-!- bogdanc [~bogdan@] has joined #shogun12:03
blackburnI am curious if it would be an entrance task :D12:03
blackburnprobably it is not12:03
heikoblackburn:  nono12:03
heikoblackburn: well12:03
heikoin fact12:03
heikothe computation is easy12:03
heikojust a norm changes compared to kernel pca12:04
heikoyou might want to check this out12:04
heikomight be an entrance task12:04
heikothe algo he showed was very similar to pca12:04
blackburnheiko: I have everything to compute largest eigenvalues/vectors12:04
heikoyou dont use the largest eigenvalues12:04
heikobut I dont know all details12:04
heikoyou are the expert on this field :)12:04
blackburnheiko: which eigenvalues I would need?12:04
heikothey might be further down the line12:05
heikolike 1-3-2912:05
blackburnfull eigendecomposition?12:05
heikoblackburn: check the paper! I dont know :)12:05
heikothis it what it boils down to12:05
blackburnyeah checking12:05
heikoI dont know how they compute12:05
blackburnheiko: yes full12:06
blackburnI don't like that!12:06
-!- bogdan__ [~bogdan@] has joined #shogun12:07
-!- bogdanc [~bogdan@] has quit [Ping timeout: 252 seconds]12:08
blackburnheiko: I hate spectral methods I have to admit12:09
heikoblackburn: haha12:09
blackburnheiko: especially ill-posed12:09
blackburnlike in LLE and Laplacian Eigenmaps12:09
heikohate is such a strong word blackburn :)12:09
blackburnthey need fine tuning12:09
blackburnheiko: w/o regularization LLE is not solvable at all12:10
blackburnso it makes more degrees of freedom12:10
sonne|workheiko: urgs I know Robert12:10
heikosonne|work: how come?12:11
sonne|workhe was at TU and we have some paper together12:11
heikooh yes12:11
heikohe mentioned the TU and KR M?ller12:11
blackburnI propose to implement techniques for dimensionality reduction and metric learning which are currently missing in scikit-learn.12:13
blackburnThere are also many kernel functions which have not been implemented.  12:13
blackburnheiko: sonne|work: ^ from the scikits mailing list12:13
sonne|workblackburn: so he implements this in scikits I guess right?12:15
sonne|workI guess a one-linear each in python12:15
blackburnsonne|work: yeah proposes to - things repeat12:15
blackburnlet me state the law of machine learning toolboxes development12:16
blackburn'every toolbox gets to the page some day'12:16
blackburnheiko: dailymail wrote about city I live12:18
n4nd0blackburn: wtf :D12:18
n4nd0blackburn: have you seen that around?12:19
heikothese Russians :)12:19
blackburnn4nd0: no didn't manage to see these things for real12:19
blackburnI have seen similar things but these are nice12:19
n4nd0heiko: they are unbelievable :)12:19
heikon4nd0: yeah12:20
blackburnheiko: n4nd0: there is an explanation why these thing happen12:20
n4nd0blackburn: and that is?12:20
blackburnn4nd0: old pipes12:20
blackburnwaterpipes I mean12:21
n4nd0do they just disappear?12:21
blackburnn4nd0: you know what happens to metal with time :)12:21
n4nd0blackburn: it gets better12:22
n4nd0like iron maiden12:23
blackburnn4nd0: good12:23
blackburnn4nd0: anyway infrastructure have not got updated since like 80s12:24
n4nd0I see12:24
n4nd0I guess it must get worse there with time providing hard weather12:25
blackburnn4nd0: at least I don't have snow now :D12:30
blackburnsonne|work said he had recently12:30
n4nd0blackburn: wow, we still have12:30
n4nd0it has been sunny every day for almost two months now12:30
n4nd0well, one and a half12:31
blackburnn4nd0: everything is wet and dirty now12:31
n4nd0still snow though12:31
n4nd0yeah, that's the disgusting part of the snow, when it ends12:31
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:31
-!- phd [] has joined #shogun12:34
-!- phoenix_ukr [] has joined #shogun12:34
heikocan you help me a bit on setting up the new dev structure here?12:36
blackburnheiko: what's up?12:37
heikojust got some time, will follow your email and then probably ask you questions :)12:37
-!- ahcorde_ [c1934745@gateway/web/freenode/ip.] has joined #shogun12:38
-!- phoenix_ukr [] has left #shogun []12:38
heikoblackburn:  so first qestion12:39
heikoupon running git flow init12:40
heikoyou say I should select the development branch for next release12:40
-!- ahcorde [c1934745@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]12:40
heikobut this thing only offers my local branches12:40
blackburndo you have 'dev' branch?12:40
heikoI just did git pull --rebase upstream master12:41
blackburnand do you have develop branch?12:41
blackburngit pull upstream develop12:41
heikoah I see, I was in master branch right?12:41
blackburnyou downloaded only master branch12:42
heikowell did not change anything12:42
blackburnnow it should fetch develop12:42
blackburnheiko: hmm why?12:43
heiko git pull upstream develop12:44
heikogit flow  init12:44
heikobut the branch is not there12:44
heikoheiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun$ git branch12:45
heiko  backup12:45
heiko  backup212:45
heiko  bugsprint12:45
heiko  gpc12:45
heiko* master12:45
blackburnheiko: hmm let me check12:46
blackburnheiko: when I did the same on my outdated copy it fetched develop branch12:46
blackburnheiko: is it your fork?12:46
heikoyes my fork12:46
heikoshould I reset everything maybe?12:47
blackburnheiko: no12:49
blackburnheiko: what happens when you git fetch upstream develop?12:49
heikoheiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun$ git fetch upstream develop12:52
heiko * branch            develop    -> FETCH_HEAD12:52
heikoheiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun$ git branch12:52
heiko  backup12:52
heiko  backup212:52
heiko  bugsprint12:52
heiko  gpc12:52
heiko* master12:52
blackburnheiko: git checkout develop?12:54
heikoheiko@heiko-ThinkPad-T420:~/Desktop/shogun/shogun$ git checkout develop12:54
heikoerror: pathspec 'develop' did not match any file(s) known to git.12:54
blackburnheiko: aham12:55
blackburnthat's an interesting case we need to resolve and *remember* how to resolve12:55
blackburnheiko: because you are working on fork we need to copy that branch somehow12:57
heikoso fork again?12:57
blackburnheiko: git checkout --track upstream/develop12:57
heikogit checkout --track upstream/develop12:57
heikofatal: git checkout: updating paths is incompatible with switching branches.12:57
heikoDid you intend to checkout 'upstream/develop' which can not be resolved as commit?12:57
blackburnheiko: give me a few minutes I'll try to do that locally12:58
heikoblackburn: I think Ill just re-fork and re-checkout12:58
heikothat usually solves my problems with git12:58
heikoand I usually get mad at it after 10 minutes of trying :D12:59
blackburnheiko: that's not the proper way12:59
blackburnheiko: oh I see you are an admin too!13:00
blackburnheiko: what do you think about that modelselection thing as gsoc idea?13:00
heikowhich one?13:00
blackburnheiko: kernel["width"].asLinearRange().from(0.1).to(10.0);13:03
heikoblackburn: I like that13:04
heikoso if we can cover all the cases (and you seemed confident in that) lets do it!13:04
heikoas gsoc you mean?13:04
blackburnheiko: yes13:04
heikowould be an easy one right?13:04
blackburnno it is not the only idea13:04
heikoor not easy but small13:04
heikowhat are the others?13:04
blackburnheiko: oh more on similar syntax13:05
blackburnheiko: features.feature(0).disable()13:05
heikoI see13:05
heikowould you mentor that?13:05
heikoor if not who would?13:05
heikoisnt that all a bit too much?13:05
blackburna bit too much of?13:05
heikofor you I mean13:06
heikoI really like the idea btw13:06
blackburnheiko: ah, don't know13:06
heikothere are many things we can do for modelselection13:06
heikosonne|work also mentioned this subsampling approach13:06
heikoto reduce costs of grid-search13:06
heikoand also there is the thing to combine C1 C213:06
blackburnheiko: I hope I manage to carry this13:07
heikoblackburn: is there any buffer? for unforeseen things?13:07
blackburnheiko: buffer like?13:07
blackburnheiko: I will be mentoring only two ideas for real13:08
blackburnso should work13:08
-!- gsomix [~Miranda@] has joined #shogun13:09
heikotime buffer13:09
heikoblackburn: well - then13:09
heikogo for it!13:09
heiko :)13:09
heikogsomix: hi!13:09
n4nd0what's up gsomix13:13
n4nd0how are you doing13:13
gsomixn4nd0: fine, I am in the process of selecting a project idea13:16
n4nd0gsomix: nice, which ones do you fancy?13:17
gsomixn4nd0: improve dimensionality reduction module... or further improvements of interfaces. not clear in my mind13:20
-!- medeeiip [~medeeiip@] has joined #shogun13:21
gsomiximprove interfaces and shogun would be great, but I don't have good ideas to suggest a new project13:23
* gsomix away13:23
n4nd0gsomix: interfaces through SWIG you mean?13:23
-!- akshay_hebbar [~akshay@] has quit [Quit: Leaving.]13:33
-!- akshay_hebbar [~akshay@] has joined #shogun13:34
-!- n4nd0 [] has quit [Ping timeout: 248 seconds]13:36
-!- bogdanc [~bogdan@] has joined #shogun13:37
-!- blackburn [] has quit [Quit: Leaving.]13:39
-!- bogdan__ [~bogdan@] has quit [Ping timeout: 276 seconds]13:41
-!- bogdanc [~bogdan@] has quit [Ping timeout: 248 seconds]13:42
-!- blackburn [] has joined #shogun13:48
-!- bogdanc [~bogdan@] has joined #shogun13:48
-!- n4nd0 [] has joined #shogun13:49
-!- n4nd0 [] has quit [Client Quit]13:49
-!- n4nd0_ [] has joined #shogun13:49
-!- n4nd0_ is now known as n4nd013:49
-!- heiko [] has quit [Quit: Leaving.]13:53
-!- bogdan__ [~bogdan@] has joined #shogun14:03
-!- akshay_hebbar1 [~akshay@] has joined #shogun14:04
-!- bogdanc [~bogdan@] has quit [Ping timeout: 256 seconds]14:04
-!- aegis1 [aegis1@] has joined #shogun14:05
-!- akshay_hebbar [~akshay@] has quit [Ping timeout: 245 seconds]14:06
n4nd0blackburn, sonne|work , heiko: do you guys know about metaoptimize?14:20
blackburnn4nd0: I don't know anything, why?14:21
n4nd0blackburn: check it out, I think it is a nice thing14:21
blackburnn4nd0: you mean their QA?14:22
blackburnI have read it quite a few times14:22
sonne|workblackburn: didn't you mention that once?14:22
n4nd0blackburn: oh yes I meant the QA forum hehe14:22
blackburnsonne|work: I don't know anything on the company but QA is nice yes14:22
sonne|workblackburn: that thing ?14:23
sonne|workblackburn: which is that link I posted14:25
blackburnsonne|work: ah yes, true14:26
sonne|workso we could just install it if we wanted that14:29
-!- ahcorde_ [c1934745@gateway/web/freenode/ip.] has quit [Quit: Page closed]14:30
blackburnsonne|work: I do not know if it makes sense14:31
n4nd0blackburn: I remember you said once something in the stackoverflow style would be more useful than the mailing list ;)14:31
blackburnn4nd0: I still think so14:33
sonne|workblackburn: but?14:33
n4nd0blackburn: when why wouldn't it make sense to install that thingy?14:33
n4nd0when -> then14:34
blackburnn4nd0: may be it would be better to use stackoverflow14:34
n4nd0mm I see14:34
blackburnn4nd0: it is essential to have people14:35
blackburnwe won't have people at standalone thing14:35
-!- akshay_hebbar [~akshay@] has joined #shogun14:35
n4nd0I think thach stackoverflow might be too big and general14:36
n4nd0but we could use the metaoptimize Q/A14:36
-!- akshay_hebbar [~akshay@] has left #shogun []14:36
-!- akshay_hebbar1 [~akshay@] has quit [Ping timeout: 245 seconds]14:37
blackburnn4nd0: tags14:37
-!- chintak [73f88294@gateway/web/freenode/ip.] has joined #shogun14:41
-!- chintak [73f88294@gateway/web/freenode/ip.] has left #shogun []14:42
-!- bogdan__ [~bogdan@] has quit [Quit: Konversation terminated!]14:48
-!- blackburn [] has quit [Read error: Connection reset by peer]14:49
-!- heiko1 [] has joined #shogun14:58
-!- medeeiip [~medeeiip@] has quit [Quit: Leaving]14:59
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun15:04
deerishidoes anyone has Andreas Ziehe email address?15:04
-!- stefan_ [] has joined #shogun15:05
n4nd0deerishi: andreas <DOT> ziehe <AT> tu-berlin <DOT> de15:05
n4nd0mmm I just got it from the ideas page15:05
n4nd0deerishi: did you already try with that one, or?15:05
deerishin4nd0: thanks. i havent tried that.15:08
n4nd0deerishi: ok, you are welcome.15:08
-!- aegis1 [aegis1@] has quit [Ping timeout: 256 seconds]15:11
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]15:12
-!- blackburn [] has joined #shogun15:13
-!- blackburn [] has quit [Quit: Leaving.]15:18
-!- stefan_ [] has quit [Quit: Konversation terminated!]15:47
-!- n4nd0 [] has quit [Quit: leaving]15:54
-!- votjak [] has joined #shogun15:59
-!- rishabh [~rishabh@] has quit [Ping timeout: 256 seconds]16:05
-!- heiko1 [] has quit [Quit: Leaving.]16:33
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun17:08
-!- gzhd79 [] has quit [Quit: gzhd79]17:08
-!- medeeiip [~medeeiip@] has joined #shogun17:10
@sonney2kI guess I made a big mistake :/17:11
-!- bogdanc [] has joined #shogun17:22
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]17:23
-!- Calvin-O [73f82292@gateway/web/freenode/ip.] has joined #shogun17:35
-!- heiko [] has joined #shogun17:41
-!- ogokal [556c1cca@gateway/web/freenode/ip.] has joined #shogun17:44
-!- votjak [] has quit [Quit: ChatZilla 0.9.90 [Firefox 20.0/20130326150557]]17:50
-!- se7en7 [~babu@2001:6b0:1:1430:21e:4fff:fec0:f4e2] has joined #shogun17:55
-!- ogokal [556c1cca@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:00
-!- rishabh [~rishabh@] has joined #shogun18:19
-!- blackburn [~blackburn@] has joined #shogun18:24
blackburnsonney2k: what is the mistake?18:24
-!- deerishi [73f88294@gateway/web/freenode/ip.] has joined #shogun18:41
-!- Calvin-O [73f82292@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:05
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has joined #shogun19:05
-!- deerishi [73f88294@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:07
-!- rishabh [~rishabh@] has quit [Ping timeout: 252 seconds]19:08
-!- n4nd0 [] has joined #shogun19:10
n4nd0what mistake sonney2k ?19:11
-!- ppletscher [] has quit [Ping timeout: 264 seconds]19:21
-!- aegis1 [aegis1@] has joined #shogun19:23
-!- lambday [0e8b614d@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:33
@sonney2kblackburn, n4nd0 I accepted to be a reviewer19:38
@sonney2kfor a couple of conferences19:38
n4nd0sonney2k: lot to read/do for it?19:38
@sonney2ktough stuff19:38
n4nd0what are the conferences?19:38
@sonney2knips and some big data thing19:38
n4nd0wow you are truly a big fish :)19:39
n4nd0out of curiosity, how does it work to be a reviewer? I mean it in the sense if it is like a normal job or get anything in exchange for the work19:40
@sonney2kyou do it for free19:40
n4nd0aham! for research's sake19:40
@sonney2kn4nd0, I used to review for basically every ML conference/journal19:40
heikosonney2k: do you know how they select people? just got an invitation for nips reviewing and am wondering why that happened. Anything else than being an author?19:41
n4nd0this is full of big fishes19:42
heikon4nd0: no not me ;)19:42
heikoI declined, no idea how to do this19:42
n4nd0sonney2k: you should go back to research, the world needs you19:42
heikon4nd0: haha :)19:42
n4nd0heiko: c'mon nips is a big thing in the ML world (I think)19:43
n4nd0so you are big fish too ;)19:43
@sonney2kheiko, it is a lot of work for nips19:43
@sonney2kn4nd0, I am no longer a real researcher19:43
heikoI even struggle to understand the papers I am interested in, so I will wait a bit more before I intend to do this19:43
@sonney2kI mean industry research is totally different from academia19:43
heikosonney2k: I heard this, but probably also interesting19:43
@sonney2kheiko, that will always be the case19:44
heikosonney2k: btw why did you change frontiers?19:44
n4nd0sonney2k: yeah, I thought that. That's we need you to come back :)19:44
@sonney2kheiko, family19:44
n4nd0That's why*19:44
heikosonney2k: so more regular working times you mean?19:44
@sonney2kheiko, a permanent job / income - that's it19:44
@sonney2kno 1-2-3 year contracts19:44
@sonney2kand then the risky proposal business19:45
heikosonney2k: yeh, I can understand that!19:45
n4nd0let's move all to Stockholm, I have the feeling the professors here have very good lives19:45
heikobeing a professor should be fine almost everywhere19:46
@sonney2kheiko, no it is tough too19:47
heikosonney2k: but in terms of contract/income?19:47
@sonney2kthen you even have people whose live depend on you19:48
@sonney2kand you need to write proposals to keep them19:48
@sonney2kso risk for yourself is low but you are responsible for others - if you screw up many will suffer19:48
n4nd0c'mon there must some nice parts being professor as well!19:49
@sonney2kn4nd0, sure there is19:50
@sonney2kyou can do whatever you want19:50
@sonney2kyou can do research19:50
@sonney2kif you have cool ideas you can get people doing these for you19:51
-!- ujjwal [c77e807e@gateway/web/freenode/ip.] has quit [Ping timeout: 245 seconds]19:51
@sonney2kpermanent job blabla19:51
n4nd0sonney2k: how long were you in research after you got your Ph.D.?19:51
@sonney2kit took a while until I got my phd19:51
@sonney2kI did ML research in academia from 1999 to 2011 IIRC19:52
@sonney2kso yes I know a lot of people 'bigshots' and they know me :)19:52
@sonney2kI hardly remember when I did my phd was it 2007 or 9?19:53
n4nd0no, I was thinking more that I think you have done lot of stuff for "not so many" years in research19:53
n4nd0but maybe I am wrong, I have not really compared in depth19:53
@sonney2kphd end of 2008 :)19:54
n4nd0I started uni back in 2008 I think :)19:55
@sonney2kI really started in 2004 (2002 I did my master, then 2 years sys admin)19:55
n4nd0sonney2k: how does it work when something big as your large-scale MKL paper happens?19:58
n4nd0is it like a happy idea one suddenly comes up with? :)19:58
@sonney2kn4nd0, yes20:00
@sonney2kand the right people20:00
@sonney2kyou need to have a problem you want to solve, someone with the skills and then someone doing guidance and sb doing the work...20:01
n4nd0I see20:02
blackburnsonney2k: nah just write 'crap' on all these papers and that's it20:15
blackburnthat's what I do when I review IEEE20:16
n4nd0wtf I am the only here who doesn't review anything?20:17
blackburnn4nd0: that's joke nobody would ever put me to review IEEE20:17
@sonney2kblackburn, that is what I intend to do ;)20:18
blackburnbeat my streak on github, 34 days!20:19
n4nd0you are the boss20:20
blackburnpromote synergy!20:21
blackburncry deeply!20:21
@sonney2kblackburn, does that remind you of sth?20:33
blackburnsonney2k: yes serialization and my code20:34
-!- gzhd79 [] has joined #shogun20:35
n4nd0I like your code metaphores20:35
blackburnone more line and kaboom20:35
blackburnn4nd0: lada!20:38
blackburnn4nd0: that's serialization code for sure20:38
@sonney2kblackburn, exactly20:39
-!- rishabh [~rishabh@] has joined #shogun20:43
-!- gzhd79 [] has quit [Quit: gzhd79]20:46
medeeiipI's going through the source,From curiosity i'm asking: i've found SG_MALLOC() wraper macro is used to call sg_generic_malloc<>(). Does this thing somehow resolves the problem with c++ template in SWIG??20:58
blackburnmedeeiip: what kind of problem?20:59
medeeiipI dont know much about SWIG. From basic I think I've to use template traits to support SWIG21:00
blackburnmedeeiip: we do not use any traits for swig21:01
medeeiipblackburn: Ok! can you tall me a couple of words about how SWIG interface is implemented for template classes e.g. SGMat21:03
blackburnmedeeiip: it is specialized21:03
blackburnSGMatrix<int32_t>, ... etc21:03
blackburnare renamed to IntMatrix21:04
@sonney2kmedeeiip, each become a new type in e.g. python21:04
blackburnand so on21:04
medeeiipOk! but about the previous question why  SG_MALLOC() macro? we could have used opencv type initialisation... Is there anything special?21:08
@sonney2kmedeeiip, how do the opencv guys do it?21:08
medeeiipthey use enum {}; CV_<bit-depth>{U|S|F}C<number_of_channels> format for initialising a Mat21:11
@sonney2kmedeeiip, we use the SG_MALLOC macro mainly to use malloc / free / realloc and so we are able to directly use memory ptrs from other languages21:16
@sonney2kas occur in python etc21:16
medeeiipis the macro is extremely necessery? From my coleg level programming exp. I think direct template based approach will be a better approach. Does it not reduces managability...... I mean hard time in debuging?21:23
n4nd0medeeiip: well in the case of SGMatrix and SGVector (SGReferencedData in general) this is managed internally21:27
n4nd0medeeiip: so one does not need to worry about it when using them21:27
n4nd0medeeiip: in any case I am not sure how using templated can make the use of the SG_MALLOC macro not necessary. I think they address different issues21:29
-!- heiko [] has quit [Quit: Leaving.]21:30
medeeiipi think the issue is when creating allocating memory i mean when calling  sg_generic_malloc<>() how the matrixtype template agrument will be determined, as we are not passing <int32_t> thing (from which the matrix type can be determined) [casual think]21:34
medeeiipas we are passing only len (argument to  sg_generic_malloc())21:35
blackburnmedeeiip: SG_MALLOC is the same as sg_generic_malloc with specified type21:36
blackburnjust shorter21:37
-!- ppletscher [] has joined #shogun21:43
-!- rishabh [~rishabh@] has quit [Ping timeout: 248 seconds]22:11
-!- FSCV [~FSCV@] has joined #shogun22:39
-!- bogdanc [] has quit [Ping timeout: 256 seconds]23:11
-!- n4nd0 [] has quit [Quit: leaving]23:12
-!- aegis1 [aegis1@] has quit []23:22
-!- harpreet [~ceo@] has joined #shogun23:58
--- Log closed Wed Apr 10 00:00:26 2013