--- Log opened Wed Aug 22 00:00:05 2012 | ||
--- Day changed Wed Aug 22 2012 | ||
n4nd0 | (I don't know how to write most today...) | 00:00 |
---|---|---|
blackburn | it appear in risk | 00:00 |
n4nd0 | well ok ... it looks like we are not going to agree today | 00:00 |
n4nd0 | my guess is that you need to explicitily look for what is the value of that max_y | 00:01 |
blackburn | risk uses violating object | 00:01 |
n4nd0 | and not say that it is going to be given by argmax_y | 00:01 |
blackburn | decision function does not | 00:01 |
blackburn | in case of correct delta it should comply with max of the second term | 00:02 |
n4nd0 | I don't think so | 00:02 |
n4nd0 | I mean, I do not agree | 00:02 |
blackburn | what can I say it works for HMM :) | 00:02 |
n4nd0 | a good model will give you for a training pair (xi,yi) | 00:03 |
blackburn | will commit things pretty soon | 00:03 |
n4nd0 | an argmax_y for xi that is close to yi | 00:03 |
n4nd0 | then the delta will be close to zero | 00:03 |
blackburn | why we want to make delta zero | 00:03 |
blackburn | if we maximize it? | 00:03 |
n4nd0 | blackburn: commit so I can take it a look | 00:03 |
blackburn | will do once I check examples | 00:04 |
yooo | rolf this yml format makes me crayz .. "The node does not represent a user object" | 00:04 |
blackburn | I hope I will manage to wake up tomorrow :D | 00:05 |
blackburn | today I was late for an hour | 00:05 |
yooo | hehe | 00:05 |
n4nd0 | hehe | 00:06 |
yooo | your commits today make me fell that one hour late didnt prevent you working on shogun during the day ^^ | 00:07 |
blackburn | I am trying to find a balance between not being fired and shoguning | 00:08 |
blackburn | n4nd0: how many iterations mosek does on that hm svm example? | 00:09 |
n4nd0 | blackburn: I am not sure right now, I think between 6-10, let me check | 00:10 |
blackburn | bmrm does 14 | 00:10 |
blackburn | takes 0.63s here | 00:10 |
n4nd0 | it takes quite longer here | 00:10 |
n4nd0 | 9 iterations | 00:12 |
CIA-52 | shogun: Sergey Lisitsyn master * r45840d7 / (12 files in 5 dirs): Refactored SO machine to not use its own features reference, fixed generic risk and bmrm result type - http://git.io/fj6dnQ | 00:12 |
n4nd0 | 26 seconds | 00:12 |
blackburn | here you go | 00:12 |
n4nd0 | wow that's a large lambda :) | 00:12 |
n4nd0 | how did you guess that value? | 00:12 |
CIA-52 | shogun: Sergey Lisitsyn master * rf253152 / (2 files in 2 dirs): Improved multiclass tree guided logistic regression - http://git.io/1RuW7Q | 00:13 |
blackburn | no idea just entered big value | 00:13 |
n4nd0 | lol | 00:13 |
blackburn | it is not the best value btw | 00:14 |
blackburn | objective tends to be pretty big this way | 00:14 |
blackburn | okay 5 millions is too big | 00:14 |
blackburn | 500 thousands tends to be overregularize too | 00:15 |
blackburn | hmm 50000 works just like 5000 | 00:15 |
blackburn | 0 is da best | 00:15 |
blackburn | will not converge | 00:15 |
blackburn | and furthermore will break you computer | 00:16 |
blackburn | I changed one already | 00:16 |
blackburn | n4nd0: with 0.01 accuracy is 0.9989 | 00:16 |
n4nd0 | blackburn: nice | 00:16 |
n4nd0 | HM-SVM? | 00:16 |
blackburn | with 1e-5 it is 0.9991 | 00:16 |
blackburn | yes | 00:16 |
blackburn | can it be 100%? | 00:17 |
n4nd0 | maybe | 00:17 |
blackburn | lets try 1e-37 | 00:17 |
blackburn | oh no | 00:17 |
blackburn | noooo | 00:17 |
blackburn | 0.9998 and crazy numbers | 00:18 |
blackburn | Fp=720255088080000357764535868940049436668877340672.000000, | 00:18 |
n4nd0 | what? | 00:18 |
blackburn | that is primal objective LOL | 00:18 |
n4nd0 | O_O | 00:18 |
n4nd0 | that looks pretty weird | 00:18 |
n4nd0 | maybe my accuracy thing is doing something wrong | 00:18 |
blackburn | no | 00:18 |
blackburn | that is ok for 1e-37 lambda | 00:19 |
blackburn | :D | 00:19 |
blackburn | n4nd0: w contain e+41 elements this way | 00:20 |
blackburn | n4nd0: okay let you compare results now :) | 00:20 |
n4nd0 | n4nd0: e+41 elements? | 00:20 |
n4nd0 | talking to myself again... | 00:20 |
blackburn | n4nd0: not number, but order | 00:20 |
n4nd0 | does the objective in bmrm minimize too? | 00:21 |
blackburn | n4nd0: do you understand what they sing about | 00:21 |
blackburn | http://www.youtube.com/watch?v=Q7xZ3cBPxIE&feature=player_detailpage | 00:21 |
blackburn | ? | 00:21 |
blackburn | :D | 00:21 |
blackburn | n4nd0: hmmmm really.. I don't know | 00:21 |
blackburn | looks like it maximiaze | 00:21 |
n4nd0 | blackburn: can you explain again why it is the argmax? :P | 00:23 |
n4nd0 | I just realized it is not feasible to check all the combinations y \in Y in HM-SVM | 00:23 |
n4nd0 | it just make sense to check all in the multiclass example | 00:23 |
blackburn | n4nd0: why loss is argmax? | 00:24 |
blackburn | because we find a most violating label for the vector | 00:24 |
n4nd0 | why max_y [ loss(yi,y) + < w, Psi(xi,y)-Psi(xi,yi) > ] | 00:24 |
n4nd0 | why the maximum is found for the argmax | 00:24 |
blackburn | ?? | 00:25 |
n4nd0 | why the maximum of the equation I have just written | 00:25 |
n4nd0 | is achieved with the argmax_y function | 00:25 |
n4nd0 | but the definition of the argmax is not to find the most violated label | 00:25 |
n4nd0 | it is to maximize <w, Psi(x,y)> | 00:25 |
blackburn | max_y [ loss(yi,y) + < w, Psi(xi,y)-Psi(xi,yi) > ] = max_y [ loss(yi,y) + < w, Psi(xi,y)> ] - <w, Psi(xi,yi) > | 00:27 |
n4nd0 | sure | 00:28 |
n4nd0 | now why | 00:28 |
n4nd0 | argmax_y [ loss(yi,y) + < w, Psi(xi,y) > ] = argmax_y < w, Psi(xi,y) > ?? | 00:29 |
blackburn | yes it is | 00:29 |
blackburn | for proper loss it is for sure | 00:29 |
n4nd0 | why? | 00:29 |
blackburn | they are both losses | 00:30 |
blackburn | first is 'label' loss | 00:30 |
blackburn | second is feature space 'loss' | 00:30 |
n4nd0 | that's not really a reason... | 00:30 |
blackburn | let me get a tea | 00:30 |
blackburn | :D | 00:30 |
n4nd0 | haha | 00:31 |
n4nd0 | we can talk about it tomorrow otherwise | 00:31 |
n4nd0 | maybe I am just stupid right now | 00:31 |
n4nd0 | and so stubborn as to see it | 00:31 |
n4nd0 | but this idea seems so simple | 00:31 |
n4nd0 | argmax_y should give a y close to yi for the input xi | 00:32 |
blackburn | I am kind of practical guy so I prefer to check if it works first :D | 00:32 |
n4nd0 | and a y close to yi implies Delta(yi,y) close to zero | 00:32 |
n4nd0 | blackburn: I cannot compile here | 00:33 |
n4nd0 | https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/structure/PrimalMosekSOSVM.cpp | 00:34 |
n4nd0 | line 30 | 00:35 |
n4nd0 | that constructor is not in | 00:35 |
n4nd0 | https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/machine/LinearStructuredOutputMachine.h | 00:35 |
blackburn | uh ok | 00:35 |
n4nd0 | did you forget to add any of the files or? | 00:36 |
blackburn | I removed features - did you forgot? :) | 00:36 |
n4nd0 | yeah I know you did that :) | 00:36 |
n4nd0 | but since the build doesn't compile | 00:37 |
-!- naywhaya1e is now known as naywhayare | 00:37 | |
blackburn | I have to do a blind fix | 00:38 |
shogun-buildbot_ | build #431 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/431 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 00:38 |
n4nd0 | ? | 00:38 |
n4nd0 | blackburn: why a blind fix? | 00:39 |
blackburn | have no mercy^W mosek | 00:39 |
n4nd0 | aaah ok | 00:39 |
n4nd0 | true fact | 00:39 |
blackburn | whicho ne? | 00:39 |
n4nd0 | I can apply that if you want | 00:39 |
n4nd0 | it looks like that was the only one | 00:40 |
blackburn | okay let you check it | 00:40 |
blackburn | n4nd0: please try now | 00:41 |
CIA-52 | shogun: Sergey Lisitsyn master * rff37c90 / (5 files in 2 dirs): Updated PrimalMosekSOSVM - http://git.io/kTbSbg | 00:41 |
blackburn | n4nd0: inequality just after A.8 in teo | 00:44 |
blackburn | this one should make you believe | 00:45 |
n4nd0 | mmm | 00:45 |
blackburn | I didn't though about that | 00:45 |
blackburn | 1 + 2 is majorizing thing for 1 | 00:45 |
n4nd0 | yes | 00:46 |
blackburn | so it is a major of risk | 00:46 |
n4nd0 | but I don't understand why the loss is being majorized too! | 00:46 |
n4nd0 | haha | 00:46 |
n4nd0 | I am becoming crazy | 00:46 |
blackburn | there is a description out there | 00:46 |
blackburn | first of all dot product is positive okay? ;) | 00:47 |
n4nd0 | ok :) | 00:47 |
blackburn | second ineq comes from max | 00:47 |
n4nd0 | ? | 00:48 |
blackburn | orr | 00:49 |
blackburn | anyway | 00:49 |
blackburn | are you convinced now? | 00:49 |
n4nd0 | I am reading the explanation around here | 00:49 |
n4nd0 | Note that (A.2) majorizes ?(y, y * ), where y * := argmaxy w, ?(x, y ) [Tsochantaridis | 00:49 |
n4nd0 | et al., 2005 | 00:49 |
n4nd0 | haha, magic copying from PDFs... | 00:50 |
blackburn | I have no idea if it has something formal | 00:51 |
blackburn | I have a gut feeling about that :D | 00:51 |
blackburn | I know that would not work in science haha | 00:51 |
CIA-52 | shogun: Sergey Lisitsyn master * r416d360 / (2 files in 2 dirs): Doc fixes - http://git.io/VRU7oQ | 00:51 |
n4nd0 | blackburn: why not? | 00:52 |
n4nd0 | do you mean you? | 00:52 |
blackburn | what do you mean now? | 00:52 |
blackburn | :) | 00:52 |
n4nd0 | < blackburn> I know that would not work in science haha | 00:52 |
blackburn | I mean no one would believe me if I said I have a gut feeling :D | 00:53 |
blackburn | about quantum mechanics or economics or anything | 00:53 |
n4nd0 | aham | 00:53 |
blackburn | okay no moar mosek | 00:53 |
blackburn | I'm satisfied now | 00:54 |
n4nd0 | :) | 00:56 |
shogun-buildbot_ | build #432 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/432 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 01:00 |
n4nd0 | blackburn: I think the same still! | 01:01 |
n4nd0 | haha | 01:01 |
n4nd0 | I believe strongly that another argmax is required | 01:02 |
blackburn | it is infeasible | 01:02 |
n4nd0 | one that includes the term Delta(yi,y) | 01:02 |
n4nd0 | it depends on the application | 01:02 |
n4nd0 | for the HM-SVM it is probably feasible | 01:02 |
blackburn | infeasible for HM, right? | 01:03 |
blackburn | okay, infeasible for CRF? | 01:03 |
n4nd0 | in the same way that argmax_y < w, Psi(x,y) > | 01:03 |
n4nd0 | blackburn: my CRF knowledge is a bit limited for this | 01:03 |
n4nd0 | but my guess is that it is feasible for CRF too | 01:03 |
n4nd0 | since they have been used in real applications | 01:04 |
blackburn | you should taklk to some SO expert | 01:04 |
n4nd0 | blackburn: do you give me some confidence at least? :D | 01:04 |
blackburn | I am pretty sure it is correct now | 01:04 |
n4nd0 | but you are just basing it in the result of the accuracy | 01:05 |
n4nd0 | take into account that the dataset is not difficult | 01:05 |
n4nd0 | the underlying distribution is not really difficult | 01:05 |
blackburn | no, I am basing it on my understanding of all these things | 01:05 |
n4nd0 | and we are providing lot of training examples | 01:05 |
n4nd0 | but look to equation A.7 in Teo | 01:06 |
blackburn | and? | 01:06 |
n4nd0 | a hint could be that they do not use the same to denot | 01:06 |
n4nd0 | denote | 01:07 |
n4nd0 | y_bar | 01:07 |
n4nd0 | and y_* | 01:07 |
n4nd0 | y_* = argmax_y < w, phi(x,y) > | 01:07 |
n4nd0 | and y_bar is the thing of A.7 | 01:07 |
blackburn | oh I have to make SO locked training work | 01:07 |
n4nd0 | ? | 01:08 |
blackburn | it is not supported now | 01:08 |
n4nd0 | anyway | 01:09 |
n4nd0 | time to sleep now | 01:09 |
blackburn | good idea | 01:10 |
n4nd0 | good night & good job :) | 01:10 |
blackburn | good night | 01:10 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 01:10 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 252 seconds] | 01:16 | |
-!- blackburn1 [~blackburn@188.168.14.172] has joined #shogun | 01:20 | |
-!- blackburn [~blackburn@37.61.181.133] has quit [Read error: Operation timed out] | 01:21 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 01:27 | |
CIA-52 | shogun: Sergey Lisitsyn master * rd6b3a7d / (9 files in 3 dirs): Refactored apply of latent machines - http://git.io/9Omc-g | 01:29 |
-!- blackburn1 [~blackburn@188.168.14.172] has quit [Quit: Leaving.] | 01:38 | |
shogun-buildbot_ | build #433 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/433 | 01:42 |
-!- yooo [575b08cb@gateway/web/freenode/ip.87.91.8.203] has quit [Ping timeout: 245 seconds] | 02:41 | |
shogun-buildbot_ | build #62 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/62 | 03:02 |
-!- av3ngr [av3ngr@nat/redhat/x-swaimdyxogxxwdlk] has joined #shogun | 05:57 | |
CIA-52 | shogun: Sergey Lisitsyn master * rcf43532 / (2 files): Fixed compilation of data generator in case of no lapack available - http://git.io/QAFu6w | 07:43 |
CIA-52 | shogun: Sergey Lisitsyn master * r40564ac / src/shogun/latent/LatentSOSVM.cpp : Added missed lambda usage in LatentSOSVM - http://git.io/jfaIwQ | 08:52 |
CIA-52 | shogun: Sergey Lisitsyn master * r157fba3 / src/interfaces/modular/modshogun_ignores.i : An attempt to ignore display methods of SG datatypes - http://git.io/t04KgA | 08:58 |
-!- uricamic [~uricamic@2001:718:2:1634:155e:7544:fba4:9878] has joined #shogun | 09:03 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 09:32 | |
n4nd0 | hey uricamic | 09:44 |
n4nd0 | how are you doing? | 09:44 |
_____________ | uricamic: hey - you should convince n4nd0 today :D | 09:49 |
n4nd0 | haha | 09:49 |
n4nd0 | _____________: uricamic and I will convince you :D | 09:50 |
n4nd0 | _____________: let me ask you | 09:53 |
n4nd0 | why do you do CDotFeatures* in 47 of PrimalMosekSOSVM.cpp | 09:53 |
_____________ | n4nd0: why not CFeatures? | 09:54 |
n4nd0 | _____________: yeah | 09:54 |
_____________ | because I didn't know what is restriction | 09:54 |
_____________ | please change if it is not correct | 09:54 |
n4nd0 | ok | 09:54 |
n4nd0 | just wanted to ensure that there was no special reason | 09:54 |
_____________ | good | 09:54 |
n4nd0 | but remember this restriction was the one that started all the refactoring | 09:55 |
n4nd0 | because in BMRM it was CDotFeatures and we needed CFeatures there | 09:55 |
_____________ | yes | 09:56 |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has joined #shogun | 10:13 | |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has quit [Client Quit] | 10:13 | |
CIA-52 | shogun: iglesias master * rb985ee1 / src/shogun/structure/PrimalMosekSOSVM.cpp : * fix compilation error with MOSEK support - http://git.io/TtNrZw | 10:23 |
CIA-52 | shogun: Sergey Lisitsyn master * r8122bd8 / src/shogun/structure/PrimalMosekSOSVM.cpp : Merge pull request #751 from iglesias/master - http://git.io/0aIxew | 10:23 |
uricamic | hi, sorry, I have been away | 10:28 |
_____________ | uricamic: we were arguing HARD yesterday | 10:28 |
_____________ | uricamic: help us to resolve that :) | 10:28 |
_____________ | uricamic: okay - the problem is | 10:29 |
uricamic | ok :) | 10:29 |
_____________ | is argmax of <w, Phi(x_i,y)> | 10:29 |
_____________ | equal to argmax of | 10:29 |
_____________ | \Delta(y_i, y) + <w, Phi(x_i,y) - Phi(x_i,y_i)> | 10:29 |
_____________ | see what I mean? | 10:29 |
n4nd0 | uricamic: put some light into here please :D | 10:30 |
uricamic | I am not sure what u mean, but wait a sec, I will figure it out :) | 10:30 |
_____________ | uricamic: I've added generic risk function | 10:30 |
_____________ | to structured model | 10:30 |
n4nd0 | uricamic: when one needs to compute the risk function | 10:30 |
_____________ | have you seen? | 10:30 |
_____________ | it works like | 10:30 |
uricamic | not yet, but I will check it now | 10:30 |
_____________ | for each feature vector: find argmax, update risk and update subgrad | 10:31 |
_____________ | but n4nd0 thinks it is wrong to find argmax without loss | 10:31 |
uricamic | yep, it is wrong | 10:31 |
n4nd0 | :) | 10:32 |
uricamic | u need to add loss there, because u are trying to find the most violated constraint there | 10:32 |
uricamic | and adapt to it | 10:32 |
uricamic | it is similar to perceptron algorithm | 10:32 |
n4nd0 | and the argmax doesn't give you the most violated | 10:32 |
uricamic | it can if u add loss term there | 10:33 |
uricamic | to all examples | 10:33 |
_____________ | okay it is a luck it works now then :) | 10:33 |
uricamic | e.g. I have been using it in one of my applications of so-svm, I have designed argmax function in matlab with the possibility to pass also the losses | 10:33 |
uricamic | _____________: and on which instance u have checked that? | 10:34 |
_____________ | HM-SVM and multiclass | 10:34 |
_____________ | n4nd0: is it feasible to implement argmax with loss? | 10:34 |
_____________ | for HM-SVM? | 10:34 |
uricamic | it should be possible | 10:35 |
_____________ | n4nd0: can you fix my code then? | 10:36 |
_____________ | what is needed is | 10:36 |
n4nd0 | _____________: it should be done modifying the Viterbi I implemented there | 10:36 |
_____________ | to add argmax_loss or so | 10:36 |
n4nd0 | uricamic: I am trying out the DualLibQPBMSOSVM and it never ends :( | 10:36 |
n4nd0 | I am doing something wrong for sure | 10:37 |
_____________ | change risk to call that, remove loss addition in risk | 10:37 |
uricamic | n4nd0: I haven't checked your code for HM-SVM yet, but wouldn't it be possible to just ass loss to the features in argmax? | 10:37 |
_____________ | and then implement argmax_loss for multiclass and HM | 10:37 |
uricamic | n4nd0: which algorithm and what lambda have u tryied? | 10:37 |
uricamic | *tried | 10:37 |
n4nd0 | uricamic: I have not set any algorithm explicitely to tell the truth | 10:39 |
uricamic | _____________: btw: I wanted to ask about the changes in my code for modular interfaces, because of java_modular failure | 10:39 |
uricamic | so it should use BMRM | 10:39 |
n4nd0 | uricamic: I thought that would use BMRM by default | 10:39 |
_____________ | I fixed that | 10:39 |
uricamic | that one has proven convergency | 10:39 |
n4nd0 | uricamic: lamda equal to 0.01 | 10:39 |
n4nd0 | it doesn't finish though | 10:39 |
uricamic | I see, that could be the problem | 10:39 |
_____________ | uricamic: do I understand right C is 1/lambda? | 10:40 |
uricamic | Like it goes through 1000 iterations and did not stop? | 10:40 |
_____________ | uricamic: ah one ask for your code too | 10:40 |
_____________ | could you please add !CSignal::cancel_computations() to while condition? | 10:40 |
n4nd0 | uricamic: 294 iterations and it just froze | 10:40 |
uricamic | I guess not just 1/lambda, it depends on the formulation of the task with C, do you have it somewhere? I guess it could be n/lambda | 10:41 |
_____________ | to make it possible to Ctrl+C that | 10:41 |
uricamic | froze? | 10:41 |
uricamic | hmm, that is strange | 10:41 |
uricamic | it might mean that the qp task is really hard to solve | 10:41 |
uricamic | and it could take really a long wjile | 10:41 |
uricamic | *while | 10:41 |
n4nd0 | I am using the so_multiclass example, in libshogun | 10:42 |
uricamic | n4nd0: it does not converge for so_multiclass ? | 10:44 |
_____________ | it was converging on my machine ehm | 10:44 |
_____________ | very fast | 10:44 |
n4nd0 | I am doing something wrong for sure then | 10:45 |
n4nd0 | but what? | 10:45 |
n4nd0 | http://pastebin.com/F6q9rrVS | 10:46 |
n4nd0 | just find bundle to look for the parts that use bmrm | 10:46 |
uricamic | try to increase lambda | 10:47 |
uricamic | a lot | 10:47 |
uricamic | because, I probably haven't mentioned it anywhere yet | 10:47 |
uricamic | but because of P3BMRM | 10:47 |
uricamic | we are using unnormalized risk | 10:47 |
uricamic | i.e. without division by number of examples | 10:47 |
uricamic | which changes the lambda a lot | 10:48 |
n4nd0 | to what number? | 10:48 |
uricamic | try something really big for start and then u can try lower | 10:48 |
n4nd0 | trying with 1000 | 10:48 |
uricamic | like 1e4 | 10:48 |
uricamic | yep, or that one | 10:48 |
uricamic | the problem is that without this unnormalized risk, P3BMRM would not be identical to BMRM | 10:49 |
n4nd0 | ok it finished now, thank you :) | 10:49 |
uricamic | it would require division of number of cp_models used depending also on the number of examples in the cp_model | 10:49 |
uricamic | n4nd0: you are welcome | 10:49 |
n4nd0 | I get this accuracy | 10:50 |
n4nd0 | SO-SVM: 82.80% | 10:50 |
n4nd0 | BMRM: 69.90% | 10:50 |
n4nd0 | is that normal? | 10:50 |
uricamic | but when using unnormalized risk, they are equal | 10:50 |
uricamic | could be, try lower lambda now :) | 10:50 |
n4nd0 | model selection for this would be nice :D | 10:51 |
uricamic | I have finished benchamrk for MNIST data yesterday, but I will probably have to change my code a bit to be able to send it in PR | 10:52 |
uricamic | I mean I need to return history of Fp, Fd and wdist from the solver, but in current form it crashes java_modular | 10:53 |
uricamic | so I guess, it will be sufficient to add getters for these to DualLibQPBMSOSVM class and everything should be fine then | 10:54 |
_____________ | uricamic: it works now | 10:56 |
_____________ | I believe it does | 10:57 |
uricamic | _____________: how come? I guess it is impossible :) | 10:57 |
_____________ | uricamic: you didn't declare constructor and it was not inherited from sgobject | 10:57 |
uricamic | because now in modular there is no support to return bmrm_return_T | 10:57 |
_____________ | so I had to declare ctor/dtor and add dummy save/load_serializable | 10:57 |
_____________ | I restored it | 10:57 |
uricamic | oh, I see | 10:58 |
_____________ | just check latest | 10:58 |
uricamic | I haven't noticed it | 10:58 |
uricamic | ok, thanks, I will check it | 10:58 |
_____________ | it was crashing python graphical example of multiclass | 10:58 |
_____________ | but now it works without any glitches | 10:58 |
_____________ | uricamic: could you implement a multiclass agrmax with loss? | 10:59 |
uricamic | _____________: I will look on it | 10:59 |
uricamic | so the design will be to have overloaded argmax one without loss and the other with it? | 11:00 |
n4nd0 | I think that is the idea | 11:02 |
n4nd0 | I'd like to skype with Nico first to ensure it | 11:02 |
_____________ | yes | 11:03 |
_____________ | I was pretty sure it would work without loss | 11:03 |
_____________ | but okay if it is not I don't mind | 11:03 |
_____________ | :) | 11:03 |
uricamic | well, I am still wondering that it was working, but it could happen :) | 11:04 |
uricamic | but I am sure, that the loss has to be there before argmax | 11:05 |
_____________ | n4nd0 thought it is only because of simple cases | 11:05 |
_____________ | I could believe so | 11:06 |
uricamic | yep, that is possible | 11:06 |
n4nd0 | I am preparing this case for multiclass classification to compare | 11:06 |
_____________ | http://25.media.tumblr.com/tumblr_m95fwdIEnn1qg3gfeo1_500.jpg hahah | 11:07 |
wiking | blackburn why u no break somebody else's code? :D | 11:07 |
n4nd0 | lol | 11:08 |
_____________ | man it is easy to apply that patch | 11:08 |
n4nd0 | the lol was for the picture :D | 11:08 |
wiking | wtf man | 11:08 |
wiking | you broke the example :) | 11:08 |
_____________ | how? | 11:08 |
wiking | i'll paste bt of gdb | 11:08 |
wiking | [DEBUG] entering LatentSVM::apply(LatentFeatures at 0x10260bd90) | 11:09 |
wiking | libc++abi.dylib: terminate called throwing an exception | 11:09 |
wiking | [ERROR] Unknown problem type | 11:09 |
wiking | Program received signal SIGABRT, Aborted. | 11:09 |
wiking | 0x00007fff8caff212 in __pthread_kill () | 11:09 |
_____________ | oh okay | 11:09 |
_____________ | that can be let | 11:09 |
_____________ | me fix it | 11:09 |
wiking | i'mjustaying | 11:10 |
_____________ | wiking: well you are not very active with that code so I didn't think I could make it hard to rebase :) | 11:10 |
wiking | :))) | 11:11 |
_____________ | all changes in your code are just | 11:11 |
_____________ | apply -> apply_latent | 11:11 |
wiking | it doesn't mean that i'm not preparing a patch | 11:11 |
wiking | yeah i know | 11:11 |
n4nd0 | uricamic: I don't manage to make so_multiclass_BMRM work properly | 11:11 |
_____________ | that is the right way to do that so just rename it and you wouldn't really need to rebase | 11:11 |
wiking | and as well the fuckup of yesterday's primalmosek :) | 11:11 |
n4nd0 | uricamic: I execute on command line ./so_multiclass_BMRM data.out 2 1000 500 1 0.01 BMRM | 11:11 |
uricamic | n4nd0: what do u mean properly? | 11:11 |
n4nd0 | and it stays in iteration 0 forever | 11:11 |
wiking | i mean the only thing is that please commit if it compiles | 11:12 |
wiking | and not when it doesn't | 11:12 |
wiking | and not having mosek is not an excuse when u edit that code ;) | 11:12 |
* wiking feeling bossy today :) | 11:12 | |
CIA-52 | shogun: Sergey Lisitsyn master * r5347c30 / src/shogun/machine/Machine.cpp : Added missed latent problem handling in apply of machine - http://git.io/8EdOQQ | 11:12 |
n4nd0 | _____________: I have to support wiking there, I am sorry :S | 11:12 |
uricamic | n4nd0: hmm, I think I have changed that example to expect data in svmllight format | 11:12 |
uricamic | but let me check that | 11:12 |
n4nd0 | uricamic: aaahm I am just using the data from the example I run | 11:13 |
uricamic | I am just compiling the latest code | 11:13 |
n4nd0 | uricamic: ok, let me know someting | 11:13 |
_____________ | well guys I have nothing to say if you are so worried with breaking your code :) | 11:13 |
n4nd0 | :) | 11:14 |
wiking | :> | 11:14 |
n4nd0 | I guess that we all want to rush a bit the things in order to be ready for release | 11:14 |
wiking | btw what has happened with fixing up the static analyzer problems? | 11:14 |
_____________ | I have been waiting for generic risk function for a while and nobody did it | 11:15 |
uricamic | hmm, it seems that I have to modified that example, cos' it refused to compile now :D | 11:15 |
_____________ | furthermore we had wrong interface with redundant features | 11:15 |
wiking | _____________: a PR + wait for comment would be nice... it really worked out for me till now quite well | 11:15 |
_____________ | so having no mosek is excuse, it is not project of mine but I did that | 11:15 |
wiking | _____________: but you've broken the code, and you couldn't see the problem as that code never compiled on your machine | 11:16 |
n4nd0 | uricamic: oh yes, you have to change the constructor | 11:16 |
wiking | DONT FUCKING BREAK THE CODE! :) | 11:16 |
n4nd0 | uricamic: just leave out features | 11:16 |
n4nd0 | uricamic: and thank _____________ :P | 11:16 |
uricamic | n4nd0: I see :D | 11:16 |
wiking | uricamic: :DDD | 11:16 |
_____________ | wiking: I know what to do better | 11:17 |
wiking | _____________: u mean u know better? :) | 11:17 |
_____________ | breaking mosek is not the issue when we are in rush to release sooner | 11:17 |
wiking | but then we have a broken release :) | 11:17 |
wiking | especially that mosek is not tested on any of the bots | 11:18 |
uricamic | n4nd0: ok it works for me now | 11:18 |
n4nd0 | _____________: we have to face that it will probably be difficult to have everything working with bundle methods for release | 11:18 |
wiking | i.e. nobody else will pick up this error just who really wants it to work :) | 11:18 |
_____________ | n4nd0 will always test it because he works with it | 11:18 |
_____________ | n4nd0: I can't see any difficult there | 11:18 |
_____________ | you said HM argmax with loss is easy? | 11:18 |
wiking | ok i was mostly joking till now | 11:18 |
n4nd0 | I didn't say it was easy | 11:19 |
n4nd0 | I am just guessing | 11:19 |
_____________ | so you can't do that? | 11:19 |
wiking | but seriously is it really hard to wait 6 hours for comments on a PR _____________ ? | 11:19 |
wiking | if yeas | 11:19 |
wiking | then mea culpa | 11:19 |
uricamic | n4nd0: so the input data are expected in svmlight format | 11:19 |
n4nd0 | _____________: I didn't say I can't | 11:19 |
_____________ | I have no fucking idea when will you appear wiking | 11:19 |
wiking | _____________: i dont have to appear to comment on a PR on githug | 11:20 |
_____________ | what to wait for? | 11:20 |
n4nd0 | _____________: I just say that maybe it cannot be done for today | 11:20 |
wiking | *github | 11:20 |
_____________ | I will do that then | 11:20 |
uricamic | i.e. on each line first number is the label and then non-zero elements of feature vectors in format idx:value | 11:20 |
wiking | especially if you add a @<github username> into the comment | 11:20 |
wiking | as that person will get an email most probably | 11:20 |
wiking | as that's the default setting of github... | 11:20 |
_____________ | wiking: what should I wait for? | 11:21 |
wiking | _____________: comment on your intended PR | 11:21 |
wiking | from the author of that actual code | 11:21 |
_____________ | which PR? | 11:21 |
_____________ | apply_latent? | 11:21 |
wiking | _____________: i'm talking about future | 11:21 |
wiking | that if u change code | 11:21 |
wiking | which u really dont feel sure about | 11:21 |
wiking | then maybe go with a PR and wait 3-6 hours | 11:21 |
_____________ | in future I would never get you back | 11:21 |
_____________ | what is unsure in that PR? | 11:22 |
wiking | and if there's no comment | 11:22 |
_____________ | it should be apply_latent but not apply | 11:22 |
_____________ | did you ever check how machine works? | 11:22 |
wiking | it's not about that given commit | 11:22 |
wiking | same goes with primal mosek per se | 11:22 |
wiking | and this is not a personal vendetta | 11:22 |
_____________ | did you check logs where I asked n4nd0 to check if it compiles? | 11:22 |
wiking | so please do not start to patronize me with "did you ever check how machine works?" | 11:23 |
wiking | i'm just suggesting | 11:23 |
wiking | that maybe it would be better... | 11:23 |
wiking | but if you know better | 11:23 |
wiking | then it's ok | 11:23 |
_____________ | should I spend time on installing mosek just to check if it compiles when I could ask n4nd0 and he don't mind to do that? | 11:23 |
wiking | sorry mate to bringing up this idea | 11:23 |
_____________ | I do not understand what you do blame me about | 11:23 |
wiking | i dont blame u | 11:24 |
wiking | i'm just saying that it would be great that if you start changing code that you haven't authored and it's not really an obvious assertation | 11:24 |
wiking | then maybe it would be great that you somehow assure that it's ok | 11:24 |
_____________ | what is not obvious? removing features or renaming apply to apply_latent? | 11:25 |
wiking | _____________: i've given there the example: not obvious = not an assertation | 11:25 |
_____________ | you claim I do that - where did I something that needs an author comment? | 11:25 |
wiking | all these commits | 11:26 |
wiking | that actually broken something | 11:26 |
_____________ | which these? | 11:26 |
wiking | see the mentioned ones above | 11:26 |
wiking | or now you really want to have sha-1 lines? | 11:26 |
wiking | i mean seriously you are now going to take this as a personal matter | 11:26 |
wiking | ? | 11:26 |
_____________ | your example should work now it is just a missed line | 11:26 |
wiking | because sorry mate i really like you so i rather just shut up | 11:27 |
_____________ | I do not understand why making it correct is so much an issue for you | 11:27 |
_____________ | yes because you are blaming me in things I didn't do | 11:27 |
_____________ | I am spending time on things I am unsure you will spend on | 11:28 |
_____________ | lately I have been fixing doc - wiking your code too - I don't know whether you would do that | 11:28 |
n4nd0 | this conversation makes me sad :( | 11:28 |
_____________ | as for your example I planned to run it later anywhay | 11:29 |
_____________ | I just wanted to make your projects more suitable before release that is coming already | 11:30 |
wiking | ok 2 more questions: a) have you commited code that broke examples and compilation of primalmosek? (and you have your answer on your line of 'blaming you') b) why do you feel attacked here, i think this really a work matter and nothing personal? but anyhow if you feel in any ways offended, then i'm honestly sorry and let's just forget this... it's really not worth it... | 11:30 |
_____________ | you are not in active development (at least it looks like that for me) | 11:30 |
_____________ | a) yes, and fixed that after n4nd0's report | 11:31 |
_____________ | b) because you are attacking me - pretty obvious | 11:31 |
_____________ | n4nd0: can you finally assure wiking I didn't break your code without your participation at all? | 11:32 |
wiking | as said in b) and earlier it wasn't meant to be attacking at all just a suggestion what would be great... | 11:32 |
_____________ | or you think too I am the breaker that makes your projects worse? | 11:32 |
wiking | dude | 11:32 |
wiking | you completely take this the wrong way | 11:33 |
wiking | nobody ever told that "I am the breaker" | 11:33 |
wiking | and that you are making the project worse | 11:33 |
wiking | you are taking this out of proportion | 11:33 |
wiking | i haven't told any of this | 11:33 |
n4nd0 | uricamic, _____________ : nice! I finally finished the test | 11:34 |
uricamic | n4nd0: great :) | 11:34 |
wiking | and wasn't even nearly implying it... | 11:34 |
n4nd0 | uricamic, _____________ : so I have compared the error when using the CMulticlassModel::risk and the error when using CStructuredModel::risk | 11:34 |
n4nd0 | the same data, same epsilon, same lambda, everything the same except from that function | 11:35 |
n4nd0 | when using CMulticlassModel::risk we get an error equal to 21.4% | 11:35 |
n4nd0 | and when using the other around 40% | 11:35 |
n4nd0 | _____________: so I finally conclude that bot argmax are not the same :) | 11:36 |
_____________ | n4nd0: I don't mind that but just do that correct | 11:36 |
uricamic | n4nd0: yep, of course they can't be :) | 11:36 |
uricamic | I am going for lunch now, will be back hopefully in less than 1 hour | 11:36 |
n4nd0 | uricamic: bye | 11:37 |
n4nd0 | _____________: c'mon man, don't say it so :( | 11:37 |
_____________ | how so? | 11:37 |
n4nd0 | never mind | 11:37 |
_____________ | I just mean if you have no time for that | 11:38 |
_____________ | just tell me | 11:38 |
_____________ | I will try to do by myself | 11:38 |
n4nd0 | look _____________, I have time for that | 11:38 |
n4nd0 | but time in my time frame | 11:38 |
n4nd0 | I may have done it by the end of this week, tested and everything | 11:38 |
n4nd0 | because first I want to consult with Nico | 11:39 |
n4nd0 | if you do not like that ..... just do it yourself | 11:39 |
_____________ | I want that to be in release I am sorry I am pushing you | 11:39 |
n4nd0 | _____________: I will do my best then | 11:39 |
_____________ | it seems to be the best new feature of shogun | 11:40 |
n4nd0 | _____________: then I will get more motivation :) | 11:40 |
_____________ | so I don't want to fuck up there | 11:40 |
n4nd0 | all right | 11:40 |
_____________ | so argmax is just viterbi in hm? | 11:41 |
n4nd0 | yes | 11:41 |
_____________ | can't you just put losses in that loss matrix? | 11:41 |
_____________ | I am newbie in HM still but just ask | 11:41 |
n4nd0 | let me check one thing in the hmsvm toolbox | 11:42 |
_____________ | it is easy to implement for multiclass | 11:43 |
CIA-52 | shogun: Soeren Sonnenburg master * r6d43e98 / src/shogun/features/DataGenerator.cpp : include lib/config.h when checking for a DEFINE_TO_BE_SET - http://git.io/cTCq-w | 11:43 |
_____________ | really easy | 11:43 |
n4nd0 | http://pastebin.com/KPkuwjta | 11:44 |
n4nd0 | take a look to that code | 11:44 |
_____________ | sonney2k: it is already included in .h - why? | 11:44 |
n4nd0 | _____________: you can see there are two calls to best_path | 11:45 |
_____________ | hah so it uses shogun's viterbi? | 11:45 |
n4nd0 | in the past yes, not in that version | 11:45 |
_____________ | lol | 11:45 |
_____________ | because our HMM is broken | 11:45 |
_____________ | and not because of me wiking | 11:45 |
n4nd0 | I think it was just to make it standalone... | 11:46 |
n4nd0 | at least they say so in the doc | 11:46 |
_____________ | n4nd0: 59-64 lines are the thing, right? | 11:47 |
n4nd0 | yes | 11:47 |
n4nd0 | and that is actually what I do in my code | 11:47 |
n4nd0 | I think that my argmax is looking for the MMV | 11:47 |
_____________ | what is MMV? | 11:48 |
n4nd0 | it is always doing the pred_path_mmv and not the pred_path | 11:48 |
n4nd0 | maximal margin violater | 11:48 |
n4nd0 | it appears in that snippet, in a comment | 11:48 |
wiking | _____________ as i said earlier i'm sorry that i've brought up this whole topic; it was completely unrespectful from me to say anything like that.... | 11:48 |
_____________ | ah right sure | 11:48 |
_____________ | wiking: you are right in what you are saying | 11:49 |
_____________ | but | 11:49 |
_____________ | I didn't break someone code if it is not obvious fix | 11:49 |
_____________ | or doc fix | 11:49 |
_____________ | or I didn't talk to author | 11:49 |
_____________ | obvious fix is your case | 11:49 |
_____________ | in case of mosek svm I was in touch with n4nd0 - no idea why he refuses to say that :D | 11:50 |
n4nd0 | because I do not agree in the way you are handling this | 11:50 |
n4nd0 | but anyway ... let's move on, this is not so important | 11:51 |
_____________ | okay one more guy against me | 11:51 |
_____________ | so | 11:51 |
wiking | as i said it was my fault so i hope we can get over this | 11:51 |
n4nd0 | _____________: you see, it is not against you or with you man | 11:51 |
wiking | JESUS | 11:51 |
_____________ | why didn't you say I should install mosek and do not touch it yesterday? | 11:51 |
wiking | please | 11:51 |
n4nd0 | _____________: I don't know how you say that if I am not even taking part .... | 11:51 |
_____________ | you said okay I will compile' | 11:51 |
_____________ | didn't you say that? | 11:51 |
-!- emrecelikten [~emre@trir-5d8000e4.pool.mediaWays.net] has joined #shogun | 11:52 | |
n4nd0 | all right guys | 11:52 |
wiking | let's just get over this and i'm sorry for bringing this up | 11:52 |
wiking | it's not worth any of this | 11:52 |
n4nd0 | sonney2k: captain, bring some peace please :) | 11:53 |
_____________ | I give up guys | 11:53 |
wiking | _____________ there's nothing to give up, keep on as is and it's all good | 11:53 |
_____________ | let you work on that separetely then if you are so hurted | 11:53 |
-!- emrecelikten [~emre@trir-5d8000e4.pool.mediaWays.net] has quit [Read error: Connection reset by peer] | 11:53 | |
wiking | _____________ nobody is hurt here.... | 11:54 |
wiking | except maybe you | 11:54 |
_____________ | with commits I have to do without much checks | 11:54 |
wiking | but that wasn't any of my intention | 11:54 |
wiking | and since it escalated into this that you think i'm saying that you are makeing bad for the project | 11:54 |
wiking | which is OBVIOUSLY not the case... it's the opposite | 11:55 |
wiking | you are pushing this project like nobody else here | 11:55 |
wiking | which is awesome | 11:55 |
n4nd0 | +1 | 11:56 |
wiking | so please just try to forget my comments and https://i.chzbgr.com/completestore/12/6/17/u5J1lVbU6UabQ5QinFqv0A2.jpg | 11:56 |
_____________ | I partly do understand your unsatisfaction | 11:57 |
_____________ | but I do not understand what made you so worried about that | 11:57 |
wiking | neither of that has happened in me :) | 11:57 |
wiking | neither unsatisfaction nor worried | 11:57 |
wiking | it was really just a comment | 11:57 |
wiking | no more no less | 11:58 |
wiking | and u do whatever you want with it | 11:58 |
wiking | just please dont take it as a personal vendetta against you and your personality | 11:58 |
_____________ | n4nd0: why didn't you say it is bad to do that blind fix? | 11:59 |
_____________ | and why do you feel it so bad if you could test that - as I asked you | 12:00 |
_____________ | I do not understand that | 12:00 |
wiking | nyipp | 12:01 |
n4nd0 | _____________: I think I told you I could fix the MOSEK thing to ensure that it will compile, but you insisted | 12:01 |
_____________ | ????? | 12:01 |
wiking | *booom* and the last 20 minutes disappeared :DDD | 12:01 |
wiking | nonono | 12:01 |
n4nd0 | I don't think it is bad to the blind fix either | 12:01 |
wiking | llalalalal | 12:01 |
wiking | lalal | 12:01 |
wiking | a | 12:01 |
_____________ | insisted what? | 12:01 |
wiking | alalalal | 12:01 |
wiking | aa | 12:01 |
n4nd0 | bad sounds quite bad | 12:01 |
wiking | alalla | 12:01 |
wiking | alala | 12:01 |
wiking | lalalala | 12:01 |
wiking | lalalalal | 12:01 |
wiking | ala | 12:01 |
wiking | ala | 12:01 |
_____________ | I wanted you to test if it compiles after removing features | 12:01 |
wiking | (do not let you speeak :DDD ) | 12:02 |
_____________ | is that wrong for you? | 12:02 |
n4nd0 | no | 12:02 |
n4nd0 | of course not | 12:02 |
_____________ | what is? | 12:02 |
n4nd0 | ? | 12:03 |
_____________ | I want to understand what is you feel wrong in handling | 12:03 |
_____________ | you said you disagree | 12:03 |
_____________ | with what? | 12:03 |
n4nd0 | how you have reacted | 12:04 |
n4nd0 | but it was nothing bad | 12:04 |
wiking | have i told that ever since a chick moved into my office my office became a geek magnet and all the guys are coming into my office chatting... :)) | 12:05 |
n4nd0 | :D | 12:05 |
wiking | i'll be a true cockblocker :) | 12:05 |
n4nd0 | haha | 12:05 |
wiking | i still need to work on some lines :) | 12:06 |
n4nd0 | _____________: seriously, I don't think you did wrong doing any fix there | 12:06 |
_____________ | then I am completely lost what I am wrong with | 12:06 |
wiking | _____________: nothing! | 12:07 |
wiking | that's the point :) | 12:07 |
wiking | and i was just practicing my trolling skills :D | 12:07 |
_____________ | I understand you are comfortable with your own code and it is rather you should do changes | 12:07 |
_____________ | but it would take a while and I want to push latest must have features before release | 12:08 |
_____________ | and make it correct | 12:08 |
n4nd0 | _____________: agree | 12:08 |
wiking | so is it over yet? | 12:08 |
n4nd0 | let's work together and have as many stable things as possible before release | 12:08 |
_____________ | I'll let you polish your code by yourselves then | 12:09 |
_____________ | wiking: so is your example worknig now? | 12:12 |
wiking | _____________ yeah awesom-ooo-ly thnx! | 12:12 |
_____________ | wiking: why latentsosvm has training but has no apply? | 12:13 |
wiking | _____________ because it's WIP | 12:13 |
wiking | it'll be no WIP by this weekend | 12:13 |
_____________ | I see | 12:14 |
wiking | we have this 'little' ongoing problem with structmodel and latentmodel and merging them | 12:14 |
wiking | as currently creating a working latentSOSVM | 12:14 |
wiking | takes too much user defined stuff | 12:14 |
wiking | you have to define at least 4 different classes | 12:14 |
wiking | and i think we shoudl take it down to 3 | 12:15 |
_____________ | ok | 12:15 |
_____________ | I am sorry I am a bit offensive but I still don't really get why did you both blame me | 12:15 |
wiking | I BLAME YOU FOR MISUNDERSTANDING ME! | 12:16 |
wiking | u r the MISUNDERSTANDER! | 12:16 |
wiking | :D | 12:16 |
wiking | do not be that one :> | 12:16 |
wiking | comeon man seriously | 12:16 |
wiking | can we get over it? | 12:16 |
_____________ | yes but I want you to understand I am not a psycho | 12:17 |
wiking | heheh i know you are not | 12:18 |
wiking | :D | 12:18 |
wiking | never was the case | 12:18 |
wiking | you are just passionate about shogun | 12:18 |
wiking | that would make you a geek but i would really like to think that you are not that type of a geek who would come into my office to try to chat with this chick :))) | 12:19 |
n4nd0 | :D | 12:19 |
_____________ | it is hard to talk about structured output machine with chick | 12:20 |
wiking | hahahah | 12:20 |
wiking | i hope they won't find these chat logs | 12:20 |
wiking | :))) | 12:20 |
wiking | or maybe i hope they'll find it :DD | 12:20 |
_____________ | I am sure no chick would ever try to read IRC LOG of MACHINE LEARNING TOOLBOX, I am pretty sure | 12:21 |
* wiking works at a very unpleasant place so that's why his implied complaning :) | 12:21 | |
_____________ | which office you are talking about? your university? | 12:21 |
wiking | heheheh she is doing phd in maths... | 12:21 |
wiking | i mean it's not soooo unprobable | 12:21 |
wiking | :))) | 12:21 |
wiking | yeah my uni :) | 12:22 |
wiking | ah yeah cool | 12:22 |
wiking | i know how i'll cockblock | 12:22 |
wiking | i'll use banners | 12:22 |
wiking | :D | 12:22 |
wiking | i'll print some banners and just raise them sometimes when i get annoyed with the jibberish :) | 12:22 |
n4nd0 | _____________: I think that the argmax I implemented does: argmax_y [ Delta(yi,y) + <w, Psi(xi,y)> ] in training | 12:23 |
n4nd0 | _____________: and argmax_y [ <w, Psi(xi,y)> ] during prediction | 12:23 |
wiking | n4nd0: it should do that yes | 12:23 |
_____________ | lol | 12:23 |
wiking | so if you have some good ideas for the banners let me know :) | 12:23 |
n4nd0 | wiking: but the definition of argmax is: y* = argmax_y <w, psi(xi,y)> or? | 12:24 |
wiking | n4nd0: well that's the thing | 12:24 |
_____________ | n4nd0: can't we assume training is for +LOSS and not-training is w/o loss? | 12:24 |
n4nd0 | _____________: do you mean for every model, or just for HM-SVM? | 12:24 |
wiking | n4nd0: it should be that in your case but i think you are trying to find the most violating ones | 12:24 |
_____________ | for all models | 12:24 |
n4nd0 | _____________: I am not sure about all models | 12:24 |
n4nd0 | _____________: let me check what I did in the multiclass one | 12:25 |
wiking | not-training w/o loss? | 12:25 |
wiking | no comply in me please explain | 12:25 |
_____________ | wiking: there are two modes in argmax | 12:25 |
_____________ | first is training one | 12:25 |
n4nd0 | _____________: in the multiclass the loss is not taking into account | 12:25 |
_____________ | n4nd0: we should change that then | 12:25 |
wiking | but classification (not-training?) is just w*psi(x,y) | 12:26 |
_____________ | n4nd0: okay wait - so HM is correct because it was correct alreadY? | 12:26 |
_____________ | wiking: yes | 12:26 |
n4nd0 | _____________: why? I don't know which one is the correct :D | 12:26 |
n4nd0 | wiking: yes | 12:26 |
_____________ | n4nd0: you said it does loss+w max | 12:26 |
wiking | so what else is not-training? :))) | 12:26 |
n4nd0 | prediction | 12:26 |
_____________ | exactly | 12:26 |
wiking | well then u don't need that | 12:26 |
wiking | :D | 12:26 |
n4nd0 | when you don't have the true label, you cannot get the loss | 12:26 |
wiking | it's just w*psi() | 12:26 |
_____________ | no, it is argmax of w*psi | 12:27 |
wiking | really? | 12:27 |
wiking | ah ok | 12:27 |
wiking | i get it | 12:27 |
_____________ | most probable label | 12:27 |
wiking | or i would get it in latent case | 12:27 |
wiking | but in simple struct as well? | 12:27 |
n4nd0 | http://www.cs.cornell.edu/people/tj/publications/tsochantaridis_etal_04a.pdf | 12:27 |
n4nd0 | look guys at algorithm 1 | 12:27 |
n4nd0 | oh shit I am getting crazy confused | 12:28 |
_____________ | what to do with it? | 12:28 |
_____________ | n4nd0: if you are optimizing with loss term already | 12:28 |
_____________ | lets just remove it from generic risk | 12:28 |
_____________ | and put loss to multiclass argmax on training=true | 12:29 |
n4nd0 | I don't understand, sorry | 12:29 |
_____________ | n4nd0: you said you compute argmax with loss already, right? | 12:29 |
n4nd0 | in the HM-SVM | 12:30 |
_____________ | yes | 12:30 |
n4nd0 | not in the multiclass | 12:30 |
_____________ | we can fix multiclass - it is not the issue | 12:30 |
_____________ | I mean currently it is being added to risk | 12:30 |
_____________ | in generic risk | 12:30 |
_____________ | https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/structure/StructuredModel.cpp line 170 | 12:31 |
_____________ | n4nd0: if we assume argmax includes loss on training we remove line 170 | 12:31 |
_____________ | and add loss to multiclass | 12:31 |
_____________ | then everything becomes correct and we are all happy | 12:31 |
n4nd0 | the thing is that I am not sure whether the argmax should include the loss | 12:32 |
n4nd0 | even if we are in training | 12:32 |
_____________ | ehm | 12:32 |
_____________ | you was sure yesterday | 12:32 |
_____________ | and uricamic said it should | 12:32 |
_____________ | I was unsure it should | 12:32 |
wiking | mmm | 12:32 |
wiking | well risk is being used in bmrm code | 12:33 |
_____________ | okay I am off for food | 12:33 |
wiking | and there the primal objective in some way will include using argmax | 12:33 |
wiking | alas the risk | 12:33 |
wiking | or? | 12:33 |
n4nd0 | _____________: yesterday I was sure that the argmax doesn't include the loss | 12:33 |
n4nd0 | _____________: that's why I said that it is not the same the argmax that was already implemented in the model than the one we need for risk | 12:34 |
wiking | n4nd0 what i dont get is that how u want to have a generic risk, where what the bmrm solves is actually labmda*||w|| + RISK | 12:34 |
wiking | so it pretty much depends on you what RISK -> p.o. is | 12:34 |
n4nd0 | but the risk can be written in terms of an argmax and the psi function | 12:35 |
wiking | well it depends what's RISK's form | 12:35 |
wiking | :) | 12:35 |
n4nd0 | so the generic risk calls these model dependent functions | 12:35 |
wiking | yeah i get it | 12:35 |
wiking | but the risk can be pretty much anything | 12:35 |
wiking | :) | 12:35 |
wiking | or? | 12:35 |
n4nd0 | R({\bf w}) = \sum_{i=1}^{m} \max_{y \in \mathcal{Y}} \left[ \ell(y_i, y) + \langle {\bf w}, \Psi(x_i, y) - \Psi(x_i, y_i) \rangle \right] | 12:36 |
wiking | until it's convex... ;) | 12:36 |
n4nd0 | I thought the risk is that, and only that | 12:36 |
wiking | this is from multiclass right/ | 12:36 |
wiking | ? | 12:36 |
n4nd0 | (paste here for better reading) | 12:36 |
n4nd0 | http://www.codecogs.com/latex/eqneditor.php | 12:36 |
wiking | i know this one | 12:37 |
n4nd0 | wiking: why just the multiclass one? | 12:37 |
wiking | but from where did you get this comment? | 12:37 |
n4nd0 | MuticlassModel.h | 12:37 |
wiking | yeah | 12:37 |
wiking | afaik it's a good risk function for multiclass | 12:37 |
n4nd0 | it is the same that appears in Teo | 12:38 |
wiking | yeah i know :) | 12:38 |
n4nd0 | page 103, equation A.2 | 12:38 |
wiking | and a lot of other places with SO | 12:38 |
n4nd0 | they don't talk there just about multiclass, right? | 12:38 |
wiking | noup | 12:38 |
wiking | wait a sec | 12:38 |
wiking | wait | 12:39 |
wiking | i'll just dig something out | 12:39 |
wiking | ok so here's another thing | 12:39 |
wiking | hahahah A FUCKING GEEK CAME TO MY ROOM BUT SINCE SHE IS NOT HERE HE DIDNT EVEN ENTER THE ROOM | 12:39 |
wiking | how fucking lame is this :))) | 12:39 |
wiking | they all want the pussy :))) | 12:39 |
wiking | so yeah n4nd0 look | 12:40 |
n4nd0 | :D | 12:40 |
wiking | http://www.cs.cornell.edu/~cnyu/papers/icml09_latentssvm.pdf | 12:40 |
wiking | page 3 eq(7) | 12:40 |
wiking | for me that is the RISK | 12:40 |
wiking | with the -C\sum_.... | 12:40 |
wiking | the whole thing after 1/2||w|| | 12:41 |
n4nd0 | aham | 12:41 |
n4nd0 | yeah, that is definitely different | 12:41 |
n4nd0 | that term is new | 12:41 |
wiking | and the Dual solver | 12:41 |
wiking | basically is intented to solve | 12:41 |
wiking | anything that is convex and in labmda*||w||^@ + RISK | 12:42 |
wiking | w/o constraints | 12:42 |
wiking | so | 12:42 |
wiking | until risk is convex | 12:42 |
wiking | you can have anything | 12:42 |
wiking | as risk | 12:42 |
n4nd0 | I see | 12:42 |
wiking | this is my understanding and that's why i thought that StructuredModel::risk should not be implemented | 12:42 |
wiking | n4nd0: but in multiclass case | 12:42 |
wiking | you are totally right though | 12:42 |
wiking | you can actually define the risk function with the help of psi and argmax | 12:43 |
wiking | but what if is your risk is something completely different but still convex | 12:43 |
wiking | see for example my case | 12:44 |
n4nd0 | but maybe this risk function we use in multiclass, and probably in HM-SVM too, may work as well | 12:44 |
wiking | and yes in my case \hat{y} and \hat{h} should be defined/given by argmax | 12:44 |
wiking | yeah it will work | 12:44 |
wiking | until you want that risk function | 12:44 |
n4nd0 | then we just make it virtual | 12:45 |
wiking | yeah | 12:45 |
wiking | but i think you should note it somewhere that this is just an example ;) | 12:45 |
n4nd0 | giving the possibility to be overridden | 12:45 |
wiking | yeah as until now | 12:45 |
n4nd0 | wiking: sure | 12:45 |
n4nd0 | in any case | 12:45 |
n4nd0 | probably it is not that great to have this generic risk for efficiency reasons | 12:46 |
wiking | but i thought that it's a bit 'narrowing down' when you already implement a risk function in an abstract class whilst your risk function could be anything | 12:46 |
wiking | anything = the PO is convex... | 12:46 |
n4nd0 | yes, I see your point and I agree with you | 12:46 |
wiking | but | 12:47 |
n4nd0 | it is not conceptually right I think | 12:47 |
wiking | there could be a protected default risk function ?:) | 12:47 |
wiking | in structuredmodel | 12:47 |
wiking | and then anybody who inherits from this class | 12:47 |
wiking | will have already a function at hand | 12:47 |
wiking | and he can just chuck that in into his risk function | 12:48 |
wiking | (call it from it) if he really doesn't know what is he doing :))) | 12:48 |
n4nd0 | sounds like a good idea | 12:48 |
wiking | or just wants to compare | 12:48 |
wiking | bmrm vs mosek | 12:48 |
wiking | btw have any of you run that ? | 12:48 |
wiking | what's the accuracy difference between mosek and bmrm | 12:49 |
wiking | on one given data set | 12:49 |
n4nd0 | not using an important data set | 12:49 |
wiking | hehe yeah | 12:49 |
n4nd0 | just randomly generated Gaussian | 12:49 |
wiking | but any dataset | 12:49 |
wiking | yeah | 12:49 |
n4nd0 | for the multiclass example | 12:49 |
wiking | and what are the numbers? | 12:49 |
n4nd0 | yeah, I have prepared that before | 12:49 |
wiking | how far r they from each other ? | 12:49 |
n4nd0 | running again right now .. | 12:50 |
n4nd0 | but I don't think these results are very trustworthy | 12:50 |
n4nd0 | I just plugged a lambda at random | 12:50 |
n4nd0 | SO-SVM: 92.60% | 12:50 |
n4nd0 | BMRM: 88.60% | 12:50 |
n4nd0 | MC: 40.00% | 12:50 |
wiking | okey | 12:51 |
wiking | so bmrm is not so far at all | 12:51 |
wiking | btw: labmda = 1/c | 12:51 |
n4nd0 | no | 12:51 |
wiking | what's your c in primal? | 12:51 |
wiking | just use lambda = 1/c | 12:51 |
n4nd0 | wiking: michal said something about that BUT also that he was not regularizing with the #examples | 12:51 |
n4nd0 | C = 1 in multiclass classification her | 12:52 |
wiking | but it's pretty cool though | 12:52 |
n4nd0 | for the PrimalMosekSOSVM | 12:52 |
wiking | and for mosek? | 12:52 |
wiking | ah ok | 12:52 |
wiking | then use lambda =1 | 12:52 |
wiking | :) | 12:52 |
n4nd0 | I am interested in knowing how faster it is | 12:52 |
wiking | n4nd0: CMath::time | 12:52 |
wiking | :P | 12:52 |
wiking | but still it's awesome that there's bmrm that can actually give similar values as mosek but it's fully opensource | 12:53 |
n4nd0 | it is pretty slow with lambda = 1 | 12:53 |
wiking | https://fbcdn-sphotos-a-a.akamaihd.net/hphotos-ak-ash3/580085_462759493754616_184994591_n.jpg | 12:54 |
wiking | :D | 12:54 |
wiking | ok i'm off to pick up something to eat | 12:54 |
n4nd0 | lol | 12:55 |
wiking | brb in 10 | 12:55 |
n4nd0 | for a new run with other data and lambda = 1000 (#examples) | 12:55 |
n4nd0 | SO-SVM: 85.30% | 12:55 |
n4nd0 | BMRM: 79.20% | 12:55 |
n4nd0 | MC: 54.10% | 12:55 |
n4nd0 | >>>> PrimalMosekSOSVM trained in 22.0718 | 13:01 |
n4nd0 | >>>> BMRM trained in 8.6328 | 13:01 |
n4nd0 | SO-SVM: 85.10% | 13:01 |
n4nd0 | BMRM: 81.90% | 13:01 |
n4nd0 | MC: 46.30% | 13:01 |
n4nd0 | http://29.media.tumblr.com/tumblr_lltzgnHi5F1qzib3wo1_400.jpg | 13:01 |
uricamic | back again, I had to read quite a big log :) | 13:07 |
uricamic | so for the argmax question | 13:07 |
uricamic | I said, that it is reasonable to have overloaded argmax function, which in training will get loss and therefore it will find the most violated constraint | 13:08 |
uricamic | and the standard argmax function which should be used for prediction of with trained W vector | 13:09 |
_____________ | uricamic: should it argmax including loss? | 13:09 |
uricamic | only in training | 13:10 |
_____________ | n4nd0: what is MC? | 13:10 |
_____________ | uricamic: sure | 13:10 |
_____________ | uricamic: my suggestion is | 13:10 |
_____________ | there is a training flag in argmax | 13:10 |
uricamic | but then for predictions it shouldn't use it | 13:10 |
uricamic | _____________: ok | 13:10 |
_____________ | in case of training it includes loss | 13:10 |
n4nd0 | _____________: linearMulticlassMachine with LibLinear | 13:10 |
_____________ | in other cases no | 13:10 |
n4nd0 | _____________: see libshogun/so_multiclass.cpp | 13:11 |
_____________ | uricamic: good for you? | 13:11 |
uricamic | _____________: ok, that should be fine then | 13:11 |
uricamic | yep | 13:11 |
_____________ | n4nd0: agree? | 13:11 |
n4nd0 | sure | 13:11 |
n4nd0 | then I had not fucking idea what the argmax should do | 13:11 |
n4nd0 | I feel stupid | 13:11 |
_____________ | ehhm I do not understand | 13:11 |
_____________ | I was wrong not you | 13:11 |
uricamic | because then it fits the design and it is easy to understand what is going on there | 13:11 |
n4nd0 | _____________: I thought the loss is not included in the argmax | 13:12 |
_____________ | ????? | 13:12 |
n4nd0 | so uricamic | 13:12 |
_____________ | what we were arguing about last night? | 13:12 |
_____________ | :D | 13:12 |
n4nd0 | uricamic: can you please check MulticlassModel.cpp | 13:12 |
uricamic | argmax should do exactly what the name says - find argument where the dot product is maximal | 13:12 |
_____________ | I was pretty sure my position is that argmax with loss is equal to argmax without | 13:12 |
_____________ | but it was wrong | 13:12 |
n4nd0 | _____________: yeah, but I am not talking about that | 13:12 |
n4nd0 | I thought that the argmax must not include the loss | 13:13 |
_____________ | to score? | 13:13 |
uricamic | and in training it has to find this maximum enhanced with losses -> i.e. find the most violated example | 13:13 |
n4nd0 | I understand | 13:13 |
uricamic | n4nd0: the version which now in shogun or somewhere else? | 13:13 |
CIA-52 | shogun: Sergey Lisitsyn master * r50ba31c / src/shogun/structure/StructuredModel.cpp : Removed redundant loss in generic risk - http://git.io/wcQu-w | 13:13 |
n4nd0 | uricamic: let me paste it for you | 13:14 |
uricamic | n4nd0: ok | 13:14 |
n4nd0 | http://pastebin.com/iK0xxBf1 | 13:14 |
n4nd0 | that is used for training currently | 13:14 |
n4nd0 | and if I understand everything properly now | 13:14 |
n4nd0 | that is kind of wrong | 13:14 |
n4nd0 | since for training | 13:14 |
n4nd0 | the score should be | 13:15 |
_____________ | n4nd0: uricamic so should we add return_value->delta in risk? | 13:15 |
n4nd0 | uricamic: what is computed in that snippet in line 8 + the loss | 13:15 |
_____________ | n4nd0: yes it is wrong | 13:15 |
_____________ | it is not wrong for *applying* | 13:15 |
_____________ | but it is wrong for *training* | 13:15 |
uricamic | n4nd0: I miss the loss there | 13:16 |
n4nd0 | uricamic: exactly | 13:16 |
_____________ | uricamic: n4nd0 only in case of training right? | 13:16 |
n4nd0 | _____________: I believe so | 13:16 |
_____________ | I can fix that | 13:17 |
uricamic | it is ok for prediction but for training u have to add loss there to find the maximum score on features with loss | 13:17 |
_____________ | or anybody is willing to? | 13:17 |
n4nd0 | _____________: I can do it | 13:17 |
_____________ | okay please do then | 13:17 |
n4nd0 | uricamic: should we expect better classification results after this change? | 13:17 |
_____________ | n4nd0: I am unsure we need delta parameter in result type | 13:17 |
n4nd0 | _____________: that was designed by Nico | 13:18 |
_____________ | if it is being added in case of training to score | 13:18 |
_____________ | we won't use it anyway | 13:18 |
uricamic | n4nd0: well I would expect that, because if you haven't counted with loss in training u actually trained different classifier | 13:18 |
_____________ | agree? | 13:18 |
n4nd0 | _____________: I am not sure if it is used somewhere right now | 13:18 |
n4nd0 | _____________: did you check? | 13:18 |
uricamic | _____________: what do u mean by resturn_value->delta? | 13:19 |
_____________ | it is being set but it won't be used by risk | 13:19 |
uricamic | *return_value | 13:19 |
_____________ | because score contains it | 13:19 |
_____________ | uricamic: https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/structure/MulticlassModel.cpp 109 | 13:19 |
_____________ | we don't need it, right? | 13:19 |
uricamic | probably not | 13:20 |
_____________ | because we don't need in when applying | 13:20 |
_____________ | it* | 13:20 |
_____________ | and it is included when training | 13:20 |
_____________ | n4nd0: agree? | 13:20 |
n4nd0 | I wouldn't remove it right away | 13:21 |
n4nd0 | I guess Nico put it there for some reason | 13:21 |
n4nd0 | but no idea | 13:21 |
_____________ | okay no need to remove it | 13:21 |
_____________ | but lets remove it's setting in multiclass model | 13:21 |
_____________ | okay I'll wait for your code :) | 13:26 |
_____________ | n4nd0: we may do that with delta though | 13:29 |
_____________ | n4nd0: we then would need to add it in risk function but subtract from maximum score | 13:29 |
_____________ | see what I mean? | 13:29 |
wiking | n4nd0: just a little bit of comment on CStructuredModel. dont u wanna pass const SGVector<float64_t>& w in argmax? | 13:29 |
wiking | n4nd0: just to reserve some space on the heap | 13:30 |
wiking | i mean stack | 13:30 |
wiking | ;) | 13:30 |
wiking | otherwise copy constructor is being called etc etc etc | 13:30 |
_____________ | please do that somebody | 13:30 |
_____________ | or I will | 13:30 |
_____________ | :D | 13:30 |
wiking | and actually | 13:31 |
wiking | in risk function | 13:31 |
_____________ | what is in risk function? | 13:31 |
wiking | don't we want to use SGVector<float64_t> instead of the simple pointers? | 13:32 |
_____________ | where? | 13:32 |
wiking | float64_t risk(float64_t* subgrad, float64_t* W, TMultipleCPinfo* info=0); | 13:32 |
wiking | float64_t* -> SGVector | 13:32 |
_____________ | why to use it? | 13:32 |
_____________ | we could but any specific reason? | 13:32 |
wiking | since we have sgvector as a wrapper for simple type arrays | 13:33 |
wiking | no other reason | 13:33 |
_____________ | wiking: to make it explicit I would keep it | 13:34 |
_____________ | it makes you think you are going to write to given memory | 13:34 |
_____________ | we need to add director model before release | 13:37 |
_____________ | I will do that | 13:37 |
_____________ | oops I am at job | 13:38 |
_____________ | damn | 13:38 |
wiking | _____________: working again? | 13:38 |
_____________ | yes starting from this week | 13:39 |
wiking | same place? | 13:39 |
_____________ | yes | 13:39 |
_____________ | netcracker | 13:39 |
_____________ | it is worth to mention that netcracker supports shogun implicitly :D | 13:39 |
wiking | \o/ | 13:40 |
wiking | put it in the release notes :> | 13:40 |
_____________ | because I do shogun instead of my job | 13:40 |
_____________ | :D | 13:40 |
wiking | (got that part ;P) | 13:41 |
wiking | n4nd0: wtf is with CLatentData CData and CStructedData? | 13:43 |
n4nd0 | wiking: you talkead with sonney2k about that right? | 13:45 |
n4nd0 | he didn't like CData I think | 13:45 |
n4nd0 | uricamic: does BMRM use argmax somewhere? | 13:45 |
_____________ | n4nd0: ?? risk | 13:46 |
uricamic | BMRM uses risk function, and risk function should use argmax inside | 13:46 |
n4nd0 | call directly to argmax | 13:46 |
wiking | n4nd0: i remember sonney2k telling us that he doesn't like the naming... that was as far as i understood :) | 13:46 |
wiking | n4nd0: that's why i'm asking to what now? | 13:46 |
wiking | _____________: dont u wanna fix this? :) | 13:47 |
n4nd0 | wiking: I don't know ... I am ok with anything regarding that to tell the truth :) | 13:47 |
wiking | and i'm TOOTALy not being ironic here :) | 13:47 |
_____________ | wiking: fix what? | 13:47 |
wiking | as much as it sounds :> | 13:47 |
wiking | _____________: so that we have this shit going on | 13:47 |
wiking | _____________: we have CLatentData (actually we had now it's CData) and CStructuredData | 13:47 |
wiking | and basically they are serving the same exact purpose | 13:48 |
n4nd0 | uricamic, _____________: so ... the accuracy doesn't change at all in this multiclass example | 13:48 |
n4nd0 | _____________, uricamic: training with argmax using or not using loss | 13:48 |
uricamic | you mean with argmax containing losses in training? | 13:48 |
wiking | _____________: so i thought it would be great to remove this redundancy and merge CLatentData and CStructuredData and just have CData | 13:48 |
uricamic | :) | 13:48 |
wiking | _____________: but sonney2k didn't like the name | 13:48 |
wiking | _____________ so we need a new name and that's it basically ;) | 13:49 |
_____________ | n4nd0: it might be for multiclass (as far as I understand) | 13:49 |
_____________ | wiking: why latent is not structureD? | 13:49 |
wiking | _____________ because it's not :) | 13:49 |
n4nd0 | uricamic: yes | 13:49 |
uricamic | n4nd0: but it could be true just for some particular example, definitely not generally | 13:49 |
n4nd0 | uricamic: ok | 13:49 |
wiking | i mean they are semantically not the same | 13:49 |
_____________ | wiking: what is different in latent? | 13:49 |
n4nd0 | uricamic: I am using also loss equal to 1 or 0 so it might normal that it doesn't make a huge difference | 13:49 |
wiking | _____________: that it's latent and not fixed contrary to structured | 13:49 |
wiking | _____________: other than that it's the exact same implementation :) | 13:49 |
uricamic | n4nd0: yep, that is another important fact | 13:50 |
wiking | _____________: it's just semantics | 13:50 |
wiking | _____________: but imho important one... | 13:50 |
uricamic | when u use e.g. L2 loss the difference should be more obvious | 13:50 |
_____________ | wiking: okay lets get back to that later | 13:51 |
n4nd0 | uricamic: it converges faster including the loss though, for the PrimalMosekSOSVM case | 13:52 |
_____________ | n4nd0: what did you change now? | 13:53 |
uricamic | n4nd0: you mean less number of iterations? | 13:53 |
n4nd0 | I am just comparing what happens when the argmax includes the loss or not | 13:53 |
n4nd0 | uricamic: I am just looking at the training time | 13:53 |
n4nd0 | uricamic: but I guess it is because the number of iterations is smaller | 13:54 |
uricamic | n4nd0: I see, but it should be reasonable since without loss it could happen that sometimes not true most violated constraint is found and therefore it needs more iterations to converge | 13:54 |
n4nd0 | uricamic: yeah, could be | 13:55 |
wiking | _____________: http://i.imgur.com/zQHv4.gif | 13:55 |
wiking | :DDDDDDDDDDDDDDDDDDDD | 13:55 |
n4nd0 | uricamic: you are an expert in this man! | 13:55 |
wiking | megalol | 13:55 |
wiking | :> | 13:55 |
uricamic | n4nd0: :D I wouldn't call myself one, but I have already written one paper about that | 13:56 |
uricamic | actually 2 but unfortunately the second one was not accepted, since someone has already done something similar -> axed by reviewers immedeately | 13:56 |
n4nd0 | too bad | 13:57 |
uricamic | it is life :) | 13:57 |
n4nd0 | uricamic: btw, I would really really prefer if the MulticlassModel::risk calls to argmax and delta loss | 13:57 |
n4nd0 | uricamic: what do you think? | 13:57 |
uricamic | n4nd0: sure, it is no problem | 13:57 |
uricamic | I will get to this hopefully soon | 13:57 |
n4nd0 | uricamic: do you want to change that? | 13:57 |
uricamic | n4nd0: yep I can, but if u want to do that feel free :) | 13:58 |
uricamic | since I have to write TR for my work which could be then referenced by shogun tutorial | 13:58 |
n4nd0 | I think I am going to stop soon of shoguning today ... | 13:59 |
n4nd0 | I should have done something for the job this morning | 13:59 |
_____________ | wiking: yes that made all russian twitter day yesterdday | 14:00 |
_____________ | n4nd0: uricamic where did you stop? | 14:01 |
_____________ | is it ok to include loss in multiclass now? | 14:01 |
_____________ | can I commit that? | 14:01 |
uricamic | _____________: you mean loss in multiclass argmax? | 14:01 |
_____________ | yes | 14:01 |
uricamic | _____________: then yep | 14:02 |
_____________ | if nobody of you are willing to do that I'll do | 14:02 |
_____________ | :) | 14:02 |
_____________ | n4nd0: job? | 14:02 |
n4nd0 | _____________: yeah | 14:03 |
n4nd0 | my professor at KTH gave me some things to do to help out in the robotics course this year too :) | 14:04 |
_____________ | teaching is great | 14:04 |
n4nd0 | he didn't tell me if he needs me in the lab this year too, for the moment I am just setting up some material | 14:05 |
n4nd0 | so to sum up | 14:06 |
n4nd0 | now both Multiclass and HM-SVM should work good with BMRM methods | 14:06 |
n4nd0 | uricamic, _____________: agree? | 14:06 |
_____________ | n4nd0: well | 14:10 |
_____________ | if HM-SVM argmaxes with loss | 14:10 |
_____________ | does it? | 14:10 |
n4nd0 | yes | 14:11 |
_____________ | okay nice | 14:11 |
_____________ | then yes, but multiclass needs to be changed | 14:12 |
_____________ | n4nd0: does structure_hmsvm_bmrm.py produce 99% still with latest changes? | 14:12 |
n4nd0 | it is changed | 14:12 |
_____________ | changed where? | 14:12 |
_____________ | by you - you mean? | 14:13 |
_____________ | if so let me merge it ;) | 14:13 |
n4nd0 | PR going | 14:13 |
n4nd0 | python structure_hmsvm_bmrm.py | 14:15 |
n4nd0 | Accuracy = 0.9989 | 14:15 |
_____________ | I ask because I removed loss from risk | 14:15 |
_____________ | I assume it is correct and score includes it | 14:16 |
_____________ | right? | 14:16 |
n4nd0 | let me think about it a moment | 14:16 |
n4nd0 | but I think one term needs to be substracted from the risk | 14:18 |
n4nd0 | psi_truth in particular | 14:18 |
_____________ | yes | 14:18 |
_____________ | multiclass has that problem | 14:18 |
n4nd0 | well not just psi truth | 14:19 |
n4nd0 | but <w, psi(xi,yi)> | 14:19 |
_____________ | yes | 14:19 |
_____________ | is it true for HMModel too? | 14:20 |
n4nd0 | I am not sure for that | 14:20 |
n4nd0 | for example | 14:20 |
n4nd0 | I don't see directly that the score in the HMMModel includes the loss | 14:20 |
n4nd0 | although it may be since a matrix used for Viterbi decoding includes the loss term... | 14:21 |
n4nd0 | I am not sure regarding this | 14:21 |
n4nd0 | I will ask Nico about it tomorrow | 14:21 |
n4nd0 | or maybe uricamic can tell us about it | 14:21 |
_____________ | n4nd0: why did you make first -INFTY? | 14:22 |
_____________ | this way you disallow it to select 0th class | 14:23 |
_____________ | if you do it this way then c=0 initially | 14:23 |
-!- av3ngr [av3ngr@nat/redhat/x-swaimdyxogxxwdlk] has quit [Quit: That's all folks!] | 14:24 | |
n4nd0 | mmm but that is weird | 14:29 |
n4nd0 | the accuracy remained the same after that fix | 14:29 |
n4nd0 | yeah of course it'd remain the same if I don't compile .... | 14:30 |
wiking | n4nd0: you talking to me? (c) ? | 14:31 |
uricamic | n4nd0: sorry, I have been away for a while, what is the question? | 14:31 |
n4nd0 | looking better now | 14:31 |
n4nd0 | wiking: talking alone I think... | 14:31 |
wiking | n4nd0: hehe it was a quote from the taxi driver | 14:32 |
_____________ | n4nd0: c=0 | 14:32 |
n4nd0 | uricamic: I need a rest now ... | 14:32 |
_____________ | please :) | 14:32 |
uricamic | n4nd0: ok :) | 14:32 |
_____________ | n4nd0: where is your power? | 14:32 |
n4nd0 | _____________: I updated the PR | 14:32 |
wiking | n4nd0: http://www.youtube.com/watch?v=bqLyTdcMLhc | 14:32 |
wiking | n4nd0: for your rest ;) | 14:32 |
n4nd0 | I lost my power | 14:33 |
CIA-52 | shogun: iglesias master * r1a87bf0 / (3 files): * fix MulticlassModel argmax, the loss must be included in training to find the MMV - http://git.io/l1c4ig | 14:33 |
CIA-52 | shogun: iglesias master * r80f82b5 / (3 files): + extension in so_multiclass example and * fix in multiclass BMRM - http://git.io/ihzbGw | 14:33 |
CIA-52 | shogun: iglesias master * rbbb3457 / src/shogun/structure/MulticlassModel.cpp : * wrong looop initialization - http://git.io/gxDe4Q | 14:33 |
CIA-52 | shogun: Sergey Lisitsyn master * r8cec5d0 / (6 files in 2 dirs): Merge pull request #752 from iglesias/master - http://git.io/GRG4TQ | 14:33 |
n4nd0 | I think i'll do the dishes during my rest instead | 14:33 |
shogun-buildbot_ | build #333 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/333 blamelist: iglesias <fernando.iglesiasg@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 14:44 |
wiking | lol | 14:45 |
wiking | shogun/io/streaming/StreamingAsciiFile.h | 14:46 |
CIA-52 | shogun: Sergey Lisitsyn master * r2f999e4 / examples/undocumented/libshogun/so_multiclass_BMRM.cpp : Update examples/undocumented/libshogun/so_multiclass_BMRM.cpp - http://git.io/PoHK1g | 14:48 |
wiking | may the ILSVRC people rot in hell | 14:48 |
_____________ | ? | 14:48 |
wiking | ILSVRC2012_val_00004220.JPEG | 14:48 |
wiking | JPEG | 14:48 |
wiking | and capital | 14:48 |
wiking | wtf :)))) | 14:48 |
_____________ | oracle loves to capitalize words | 14:49 |
wiking | and putting tars in tar is really a smart idea... | 14:49 |
_____________ | putting a tar archiver into a tar is way smarter | 14:49 |
wiking | heheh | 14:49 |
wiking | yeah that'dbe | 14:49 |
wiking | but still now i'm running low actually on space on the cluster :) | 14:50 |
wiking | i mean not the big hdfs | 14:50 |
wiking | but the local little /home fs | 14:50 |
_____________ | how big it is untared? | 14:50 |
wiking | well i guess it's the same size | 14:50 |
_____________ | yeah makes sense | 14:50 |
wiking | so i would need another +138 gigs | 14:50 |
wiking | but i think i've found a way around here :) | 14:50 |
wiking | since the / has 800 gigs free on one node | 14:51 |
wiking | that is part of the free space of hdfs | 14:51 |
wiking | buuut i could use that for temp time | 14:51 |
wiking | they put 1000 tars within that big tar | 14:53 |
wiking | i think the 1000 tars is the different classes | 14:53 |
wiking | lol | 14:55 |
wiking | this is nice | 14:55 |
wiking | actually they've provided a matlab file with all the dsift features within a class | 14:55 |
* wiking wonders if it's all the images in the class | 14:55 | |
_____________ | dense sift? it must be way too big to be large scale | 14:55 |
wiking | it is but still | 14:56 |
wiking | it's there | 14:56 |
_____________ | however what is the other way? | 14:57 |
_____________ | matching? | 14:57 |
_____________ | or bow like yeah | 14:57 |
wiking | mmm lol yeah | 14:57 |
wiking | we have already all the sifts if we want it :) | 14:57 |
wiking | marvolous | 14:57 |
_____________ | wiking: what is deadline for ilsvrc? | 14:58 |
wiking | 23rd of sept | 14:58 |
_____________ | I see | 14:58 |
-!- _____________ [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Ping timeout: 245 seconds] | 15:07 | |
shogun-buildbot_ | build #351 of deb2 - static_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/351 blamelist: iglesias <fernando.iglesiasg@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 15:20 |
shogun-buildbot_ | build #444 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/444 blamelist: iglesias <fernando.iglesiasg@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 15:45 |
shogun-buildbot_ | build #334 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/334 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 15:50 |
-!- alexlovesdata [~binder@goldenezahl.ml.tu-berlin.de] has joined #shogun | 15:58 | |
shogun-buildbot_ | build #445 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/445 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 16:10 |
-!- blackburn [~blackburn@188.168.13.28] has joined #shogun | 16:12 | |
shogun-buildbot_ | build #352 of deb2 - static_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/352 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 16:12 |
blackburn | time to fix things | 16:13 |
uricamic | blackburn: btw I have checked IRC log, BMRM cannot be used with lambda=0 | 16:16 |
blackburn | uricamic: sure | 16:16 |
blackburn | that's clear | 16:16 |
uricamic | and also, it should not be even close to 0 | 16:16 |
blackburn | well it converges with small lambda | 16:16 |
blackburn | but it makes no sense | 16:17 |
blackburn | it is just overfitting | 16:17 |
uricamic | blackburn: yep, I just wanted to point that out just for sure :) | 16:17 |
blackburn | the only reason I did that | 16:17 |
blackburn | to try to fit model perfectly | 16:17 |
blackburn | on training data | 16:17 |
blackburn | with no regularization at all | 16:17 |
uricamic | well, it can happen that for some instance and features the regularization term is not needed, by then it is not good idea to use BMRM | 16:17 |
uricamic | blackburn: I see, but BMRM is not designed for such things, lambda has to be >0 | 16:18 |
blackburn | uricamic: btw multiclass model is not typical in bmrm, right? | 16:18 |
blackburn | I mean it is L2 of whole w | 16:18 |
uricamic | I am not sure that I understand the question now | 16:19 |
blackburn | ahhh nevermind | 16:19 |
blackburn | it is ||w||^2 so it is the same | 16:19 |
blackburn | I meant that in crammer-singer or weston one | 16:20 |
blackburn | \sum ||w|| is used | 16:20 |
uricamic | you mean the way how it is implemented in shogun now? | 16:20 |
blackburn | but it is the same in cases of ||w||^2 | 16:20 |
uricamic | I see | 16:20 |
blackburn | read ^ as \sum ||w_i||^2 | 16:20 |
uricamic | yep | 16:21 |
blackburn | uricamic: so are we ok now with multiclass bmrm? | 16:22 |
blackburn | uricamic: btw what is the difference between all these solvers? | 16:22 |
uricamic | I will check multiclass bmrm, soon, but I guess it should be ok now | 16:23 |
uricamic | the difference | 16:23 |
uricamic | BMRM is the standard method as described by Teo et al. | 16:23 |
uricamic | only the inactive cutting plane (ICP) removal strategy is added there | 16:24 |
blackburn | how one should choose a solver? | 16:24 |
uricamic | PPBMRM is BMRM enhanced by prox-term | 16:24 |
blackburn | kind of initial guess? | 16:25 |
uricamic | which is our method how to enforce possibility of some initial solution by restricting the L2 distance between consecutive W_t | 16:25 |
blackburn | our - your and vojta? | 16:26 |
uricamic | and the last one P3BMRM is the same as PPBMRM plus it can use multiple cutting plane models, it is similar to Joachim's nslack | 16:26 |
uricamic | yep | 16:26 |
uricamic | when cp_models = 1 in P3BMRM then it is reduced to PPBMRM | 16:27 |
uricamic | and when you set K = 0 PPBMRM should reduce back to BMRM (but I haven't checked this particular one yet) | 16:27 |
blackburn | we have to make your example support default values | 16:27 |
uricamic | I mean in Shogun's implementation | 16:27 |
uricamic | you mean cpp example? | 16:28 |
blackburn | yes | 16:28 |
uricamic | ok, I put it there just for my convenience to test all algorithms quickly | 16:28 |
blackburn | I don't mind command line commands actually | 16:28 |
blackburn | but we need it able to run as is | 16:29 |
uricamic | like ./so_multiclass_BMRM and that's all? | 16:29 |
blackburn | yes | 16:29 |
blackburn | Makefile does that | 16:29 |
blackburn | when testing | 16:29 |
uricamic | i.e. to use randomly generated data and default settings | 16:29 |
uricamic | ok | 16:29 |
blackburn | yeah that would work I think | 16:29 |
blackburn | can you do that next time? | 16:30 |
uricamic | sure | 16:30 |
blackburn | the earlier the better because currently test is broken: http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/352/steps/test%20libshogun/logs/stdio | 16:30 |
uricamic | I think I will do this either in the evening or tomorrow | 16:30 |
uricamic | ahh, ok | 16:30 |
uricamic | so back to the solvers, if u are tuning the optimal lambda, it is nice to have some reasonable range of lambda's and in the cycle use PPBMRM or P3BMRM, since then it will use solution from the last lambda as initial one and converges much faster, especially for lower values of lambda | 16:33 |
uricamic | to put it in numbers, on MNIST data (10 classes, 784 feature dimension), 60k training examples | 16:34 |
uricamic | the difference between BMRM and P3BMRM on lambda=1 | 16:34 |
blackburn | kind of hot start? | 16:35 |
uricamic | yep | 16:35 |
blackburn | nice | 16:35 |
uricamic | was from 2.01 hours to 2.09 hours for P3BMRM | 16:35 |
uricamic | shit | 16:35 |
uricamic | 0.209 for P3BMRM :D | 16:35 |
uricamic | otherwise it would be so impressive :D | 16:35 |
uricamic | oh, again, wouldn't :D | 16:36 |
blackburn | 2 hrs? | 16:36 |
blackburn | :D | 16:36 |
uricamic | in iterations it was from 5932 iterations to 408 | 16:36 |
blackburn | uricamic: 0.209h sounds like it is pretty fast, something near to liblinear | 16:37 |
uricamic | and the other thing about P3BMRM is that it could be quite easily written to use more cores | 16:37 |
uricamic | blackburn: I haven't tried liblinear yet :) | 16:38 |
uricamic | but this multicore thing has probably reason only when computation of the risk takes a long time | 16:39 |
blackburn | I see | 16:39 |
uricamic | so in multiclass it would probably not give any speedup at all | 16:39 |
blackburn | that's clear | 16:39 |
uricamic | another thing is that both PPBMRM and P3BMRM tends to converge to more precise solution than BMRM because of our alpha setting strategy and termination condition | 16:40 |
blackburn | last two weeks I've been trying to employ L1/L2 tree regularization | 16:41 |
blackburn | no success :( | 16:41 |
-!- blackburn1 [~blackburn@188.168.4.46] has joined #shogun | 16:45 | |
-!- blackburn [~blackburn@188.168.13.28] has quit [Ping timeout: 250 seconds] | 16:47 | |
blackburn1 | at some point users must claim us as an OSS having best support ever | 16:50 |
blackburn1 | I will eat them if not | 16:55 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 17:02 | |
-!- blackburn [~blackburn@83.234.54.112] has joined #shogun | 17:03 | |
-!- blackburn1 [~blackburn@188.168.4.46] has quit [Ping timeout: 244 seconds] | 17:04 | |
-!- blackburn1 [~blackburn@188.168.2.65] has joined #shogun | 17:18 | |
CIA-52 | shogun: Sergey Lisitsyn master * r90d4424 / (9 files in 4 dirs): Refactored multiclass machines to support C parameter selection - http://git.io/rfAykA | 17:20 |
-!- blackburn [~blackburn@83.234.54.112] has quit [Ping timeout: 252 seconds] | 17:21 | |
-!- uricamic [~uricamic@2001:718:2:1634:155e:7544:fba4:9878] has quit [Quit: Leaving.] | 17:22 | |
shogun-buildbot_ | build #335 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/335 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 17:27 |
wiking | ahahhahahahahaha | 17:41 |
wiking | "otherwise my concern would be -- where is the unittest? logic-wise | 17:41 |
wiking | it should indeed accomplish the desired goal but imho test would help to | 17:41 |
wiking | 'guarantee' that ;) | 17:41 |
wiking | " | 17:41 |
wiking | :DDDD | 17:41 |
wiking | people are damanind :) | 17:42 |
wiking | demanding :P | 17:42 |
wiking | so yeah where's the unit test :> | 17:42 |
shogun-buildbot_ | build #446 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/446 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 17:55 |
shogun-buildbot_ | build #353 of deb2 - static_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/353 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 17:57 |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has joined #shogun | 18:02 | |
yoo | hi all | 18:02 |
wiking | yo | 18:13 |
-!- yoo [2eda6d52@gateway/web/freenode/ip.46.218.109.82] has quit [Quit: Page closed] | 18:28 | |
-!- emrecelikten [~emre@trir-5d8000e4.pool.mediaWays.net] has joined #shogun | 18:34 | |
-!- Netsplit *.net <-> *.split quits: shogun-buildbot_, yoh, audy, emrecelikten, @sonney2k, naywhayare, sr___, wiking, zxtx, CIA-52, (+1 more, use /NETSPLIT to show all of them) | 19:38 | |
-!- Netsplit over, joins: CIA-52, sonney2k, audy, shogun-buildbot_, sr___, wiking | 19:41 | |
-!- naywhaya1e [~ryan@spoon.lugatgt.org] has joined #shogun | 19:41 | |
-!- Netsplit over, joins: yoh, emrecelikten, blackburn1, zxtx | 19:41 | |
-!- ServerMode/#shogun [+o sonney2k] by moorcock.freenode.net | 19:41 | |
shogun-buildbot_ | build #336 of bsd1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/336 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 19:43 |
-!- Netsplit *.net <-> *.split quits: yoh | 19:45 | |
-!- Netsplit over, joins: yoh | 19:45 | |
wiking | blackburn1: we were in a split | 19:45 |
blackburn1 | hehe | 19:45 |
wiking | so any ideas there? | 19:46 |
blackburn1 | about?? | 19:47 |
wiking | ah the func arg | 19:47 |
wiking | have you got my two example cases for diff dims? | 19:47 |
wiking | http://pastebin.com/VhXVs7qF and http://pastebin.com/p3EFtZdf | 19:47 |
wiking | are two actual applications | 19:47 |
blackburn1 | this is *not* different dims case | 19:48 |
blackburn1 | just like in multiclass it is just sparse | 19:48 |
wiking | blackburn1: noup http://pastebin.com/p3EFtZdf | 19:48 |
wiking | check this one | 19:48 |
wiking | score += sm->w[cur_class*sparm->size_hog+i+1]*hog[i]; | 19:48 |
wiking | w = num_class*sparm->size_hog | 19:49 |
blackburn1 | it is just multiclass hog | 19:49 |
wiking | i mean the dimension of w | 19:49 |
wiking | otherwise the indexing would not work | 19:49 |
blackburn1 | dimension is W is always equal to dimension of Psi, but Psi can be sparse | 19:49 |
blackburn1 | that is the case | 19:49 |
blackburn1 | and multiclass is the case | 19:49 |
wiking | ano | 19:49 |
wiking | no | 19:50 |
wiking | ok | 19:50 |
wiking | u r right | 19:50 |
blackburn1 | dimension of Psi here is not equal to dimension of hog | 19:50 |
blackburn1 | it is dim of hog * n_classes | 19:50 |
blackburn1 | but everything out given class is 0 | 19:51 |
blackburn1 | so it is not computed at all | 19:51 |
wiking | my bad | 19:51 |
wiking | sorry | 19:51 |
blackburn1 | actually if you have that code HOGSO Model is straightforward | 19:53 |
wiking | mmm now i'm in trouble... | 19:55 |
wiking | i need to calculate this for all examples: argmax_{(ybar,hbar)} [<w,psi(x,ybar,hbar)> + loss(y,ybar,hbar)]. | 19:55 |
wiking | and i want to keep the COFFIN fw | 19:56 |
blackburn1 | line 25 is a dense dot | 19:56 |
blackburn1 | wiking: it is argmax alreadyy | 19:58 |
wiking | which? | 19:58 |
wiking | i mean now i'm writing a generic implementation i'm not talking about those pastebins | 19:58 |
wiking | it's just that currnetly i have 1 method for psi | 19:58 |
wiking | is that CDotFeatures* get_psi....() | 19:58 |
wiking | i guess i could generate all the different CDotFeatures and generate a matrix of the w,psi(x,ybar,hbar)> + loss(y,ybar,hbar) values | 20:01 |
wiking | and do there a max value search for each row/column (depending on who do i store [w,psi(x,ybar,hbar)> + loss(y,ybar,hbar)]) and then sum it up | 20:02 |
wiking | :) | 20:02 |
wiking | it's just a lot of memory waste... | 20:02 |
wiking | blackburn1: input? | 20:03 |
blackburn1 | yeah sounds inefficient | 20:04 |
wiking | but i dont have something like SGVector<float64_t> get_psi(x,y,h) | 20:04 |
wiking | as sparse vector wouldn't work then | 20:05 |
blackburn1 | why don't you have it? | 20:05 |
wiking | because i couldn't have sparse vector | 20:05 |
wiking | this is why i threw out the first place the get_psi(x,y,h) and have only CDotFeatures* get_psi_vectors() | 20:06 |
blackburn1 | I see | 20:11 |
shogun-buildbot_ | build #447 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/447 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 20:12 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 20:20 | |
n4nd0 | wiking: I just read the the thread with the problem with risk function args :D | 20:22 |
wiking | :D | 20:28 |
-!- blackburn [~blackburn@83.234.54.163] has joined #shogun | 20:42 | |
-!- blackburn1 [~blackburn@188.168.2.65] has quit [Ping timeout: 248 seconds] | 20:43 | |
-!- sr___ [u5548@gateway/web/irccloud.com/x-fgcntzhdfnowhati] has quit [Read error: Connection reset by peer] | 21:03 | |
-!- sr___ [u5548@gateway/web/irccloud.com/x-ewmhdcvbknaunxsl] has joined #shogun | 21:03 | |
-!- blackburn [~blackburn@83.234.54.163] has quit [Quit: Leaving.] | 21:15 | |
-!- naywhaya1e is now known as naywhayare | 21:38 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Read error: Connection reset by peer] | 21:56 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 21:58 | |
n4nd0 | any clue why that crash is happening in Arthur's question? | 22:03 |
n4nd0 | mailing list | 22:03 |
wiking | n4nd0: swig version could be a possible problem | 22:11 |
wiking | i mean i have 2.0.8 | 22:11 |
wiking | 1.3.29 seems a bit old | 22:11 |
n4nd0 | wiking: that's true | 22:12 |
n4nd0 | wiking: should I answer him or do you want to do it? | 22:13 |
wiking | mmm i think that should be tested though | 22:13 |
wiking | and if that's the case then we should find the minimum swig version shogun requires to be able to compile the modular interfaces | 22:14 |
wiking | the problem is that i have 0 knowledge about swig | 22:15 |
wiking | i think gsomix would be the most competent to say anything about this | 22:15 |
n4nd0 | wiking: I remember once my compilation failed and sonney2k told me to update swig | 22:16 |
n4nd0 | it worked after that | 22:16 |
n4nd0 | I will just tell him in case it can solve his problem | 22:16 |
@sonney2k | n4nd0, wiking yes we need newer swig - but IIRC configure should fail if swig < 2.0.4 | 22:17 |
wiking | yeah the problem is i think that he would need to manually install a new version of swig as he now uses the standard distrib supplied swig | 22:17 |
wiking | ooh yeah | 22:19 |
n4nd0 | I could do it via apt-get easily | 22:19 |
wiking | _swig_version=`${SWIG} -version 2>/dev/null | grep Version | cut -f 3 -d ' '` | 22:19 |
wiking | if assert_version swig $_swig $_swig_version 2.0.4; then | 22:19 |
wiking | echores "$_swig_version" | 22:19 |
wiking | _swig=yes | 22:19 |
wiking | n4nd0: it's RH :) | 22:19 |
n4nd0 | wiking: RH? | 22:19 |
wiking | redhat | 22:20 |
wiking | according to his email he uses redhat => no apt-get ;P | 22:20 |
wiking | lol his configure log is | 22:21 |
wiking | ============ Checking for SWIG ============ | 22:21 |
wiking | Result is: 1.3.29 | 22:21 |
wiking | ########################################## | 22:21 |
wiking | and the assert_version doesn't fail | 22:21 |
-!- blackburn [~blackburn@188.168.13.61] has joined #shogun | 22:21 | |
wiking | so there's something wrong with that if line | 22:22 |
blackburn | discussing swig issue? I have to note he is using 1.1.0 | 22:23 |
blackburn | I do not remember if it requires >2.0 but probably it does | 22:23 |
wiking | blackburn: just joined 1 minute later | 22:23 |
wiking | blackburn: yeah accoding to ./configure script it should actually fail because it's not 2.0.4 or later | 22:24 |
wiking | but for some reason that if branch keeps being true | 22:24 |
blackburn | he seems to be a hardcore guy | 22:24 |
blackburn | using RHEL | 22:24 |
blackburn | n4nd0: did you graduate from kindness school? :) | 22:26 |
n4nd0 | blackburn: haha why so? | 22:26 |
wiking | doh | 22:27 |
n4nd0 | it is free to be kind, not expensive :D | 22:27 |
blackburn | :D | 22:27 |
wiking | if i set _swig_version by hand the configure script fails | 22:27 |
wiking | aaah | 22:28 |
wiking | i think he is using a rather old shogun version | 22:28 |
wiking | where this check is not yet present in ./configure script | 22:28 |
wiking | as Result is: gcc 4.1.2, ok | 22:28 |
wiking | and afaik that should fail as well with < 4.3 | 22:28 |
wiking | i wonder why they want to use shogun @ ernst & young ;P | 22:36 |
blackburn | hey really | 22:37 |
blackburn | what the heck | 22:37 |
wiking | maybe for tax auditing :DDD | 22:37 |
blackburn | why anybody in ernst & young would need shogun | 22:37 |
wiking | i guess some classification stuff on tax auditing data ;P | 22:37 |
blackburn | lol | 22:37 |
wiking | would be great to hear more what's his idea for application | 22:38 |
blackburn | hey we must get more E&Y users | 22:38 |
blackburn | I bet they have money | 22:38 |
blackburn | I don't mind to become rich :D | 22:38 |
wiking | hahahahahha | 22:38 |
wiking | yeah you should do a shogun spin-off ;P | 22:39 |
wiking | and make money of shogun support :))))) | 22:39 |
blackburn | n4nd0: did you notice | 22:39 |
wiking | i mean this guy should just fucking update his linux distrib | 22:39 |
blackburn | his name is sean | 22:39 |
blackburn | but email is arthur edge | 22:39 |
wiking | ah lol | 22:40 |
wiking | wtf? | 22:40 |
wiking | maybe he is disguising himself ;P | 22:40 |
n4nd0 | haha yeah I noticed | 22:40 |
wiking | how old is RHEL 5.8 actually? | 22:40 |
blackburn | I like footer part | 22:40 |
n4nd0 | I surprised myself when I thought, I am going to change the signature just in case it is not the same as the mail | 22:40 |
blackburn | wiking: 2010 probably | 22:42 |
n4nd0 | to check* not to change | 22:43 |
wiking | duuude | 22:45 |
wiking | 2012!!!! | 22:45 |
blackburn | Any U.S. tax advice contained in the body of this e-mail was not intended or written to be used, and cannot be used, by the recipient for the purpose of avoiding penalties that may be imposed under the Internal Revenue Code or applicable state or local tax law provisions. | 22:45 |
wiking | WTF?!?!?!?!? | 22:46 |
blackburn | wiking: really? | 22:46 |
blackburn | latest is 6.3 IIRC | 22:47 |
wiking | blackburn: https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Enterprise_Linux/5/html/5.8_Release_Notes/index.html | 22:47 |
blackburn | I had to work with 5.3 yesterday on dev server | 22:48 |
wiking | blackburn: http://www.h-online.com/open/features/What-s-new-in-Red-Hat-Enterprise-Linux-5-8-1439890.html | 22:48 |
wiking | and it's a 2012 february news | 22:48 |
wiking | wtf | 22:48 |
wiking | redhat ships with fucking gcc 4.1.x in 2012? | 22:48 |
blackburn | it is ENTERPRISE redhat | 22:49 |
blackburn | i.e. | 22:49 |
blackburn | slow | 22:49 |
blackburn | very slow | 22:49 |
blackburn | very very slow | 22:49 |
wiking | so it must be filled with acient stuff? :) | 22:49 |
blackburn | exactly | 22:49 |
blackburn | very very slow developing | 22:49 |
wiking | i guess they are following: 'the older the better' | 22:49 |
wiking | ;) | 22:49 |
blackburn | they would use win95 | 22:49 |
blackburn | if m$ did support it still | 22:49 |
wiking | ah lol | 22:49 |
wiking | there's gcc44 package | 22:49 |
wiking | so he should be ok | 22:49 |
blackburn | wiking: I know a company which forces my company to use ClearCase | 22:49 |
blackburn | heard about that? | 22:49 |
wiking | hahahah YES | 22:49 |
wiking | that's like fucking 2000 | 22:49 |
blackburn | used? | 22:49 |
wiking | :D | 22:49 |
wiking | yes unfortunately | 22:49 |
blackburn | that's like | 22:49 |
blackburn | that's like nothing on the earth | 22:49 |
blackburn | it is pure love | 22:49 |
wiking | yeah it's a pure beauty | 22:49 |
wiking | :)))hahahhah | 22:49 |
blackburn | it is when you lose your mind of happiness | 22:50 |
blackburn | :D | 22:50 |
blackburn | but serious, did they really think that dance is a good way to develop | 22:50 |
blackburn | you have to create an integration stream to view sources | 22:50 |
wiking | hahahah | 22:50 |
wiking | NOOOOOO | 22:50 |
blackburn | then you add activity there you specify what do you want | 22:50 |
wiking | i dont want to hear about that shit! | 22:50 |
wiking | i had physical pain | 22:50 |
blackburn | and then you need to create a development stream | 22:50 |
wiking | when i had to use it | 22:50 |
blackburn | and join project with it | 22:50 |
wiking | yeah | 22:50 |
blackburn | then checkout or hijack the file where you want to remove 2 lines | 22:50 |
blackburn | commit (or how is it called out there) | 22:50 |
wiking | i mean seriously why would u not use git+github in 2012? | 22:50 |
blackburn | and write down 40 pages description of the commit | 22:50 |
blackburn | we are using svn | 22:50 |
wiking | blackburn: and i hope u use git-svn ;) | 22:50 |
blackburn | but one project uses clearcase still | 22:50 |
wiking | blackburn: just to make them cry :)))) | 22:50 |
blackburn | that is restricted and I didn't say that but it is TELUS | 22:50 |
blackburn | :D | 22:50 |
blackburn | wiking: you are the only one but me developer of shogun | 22:50 |
blackburn | who have a sexual contact with CC | 22:51 |
blackburn | had* | 22:51 |
blackburn | I have still though | 22:51 |
wiking | :DDD | 22:51 |
wiking | fucking clearcase | 22:51 |
wiking | i know a guy | 22:51 |
wiking | !!! | 22:51 |
wiking | he just had to start to use it | 22:51 |
wiking | :DDDD | 22:51 |
blackburn | in 2012 | 22:51 |
wiking | yep yep | 22:51 |
blackburn | sounds like torquemada is back | 22:51 |
wiking | 2 weeks ago or something | 22:51 |
blackburn | and he is full of will to punish somebody | 22:51 |
wiking | yep | 22:51 |
wiking | he seemed desparate on facebook | 22:51 |
wiking | :P | 22:51 |
blackburn | that's quite normal to get a depression after using CC for a while | 22:52 |
blackburn | I am pretty sure there was a case of commiting suicide after trying to get that thing to get your changes | 22:54 |
wiking | hehehe | 22:54 |
wiking | i think a git-cc project would never be possible | 22:54 |
blackburn | I am not using git-svn btw | 22:54 |
wiking | LOOOOL | 22:54 |
wiking | i was proved WRONG | 22:54 |
wiking | *kabooom* | 22:54 |
wiking | https://github.com/charleso/git-cc | 22:54 |
wiking | :DDDD | 22:54 |
blackburn | I actually feel comfortable with svn as well | 22:55 |
wiking | "Warning | 22:55 |
wiking | ======= | 22:55 |
wiking | I wrote this purely for fun and to see if I could stop use Clearcase at work | 22:55 |
wiking | once and for all." | 22:55 |
blackburn | for the thing we develop it is ok | 22:55 |
wiking | :DDDDDDDDDDDDDD | 22:55 |
wiking | blackburn: yeah i just don't like that u actually cannot commit a code in svn w/o having internet connection | 22:55 |
wiking | "Actually what I would love to see | 22:56 |
wiking | more is for Clearcase to die" | 22:56 |
blackburn | wiking: that is ok when you have pretty reliable internet connection at your job | 22:56 |
wiking | lol this guy did feel the pain | 22:56 |
blackburn | I share his pain | 22:56 |
blackburn | okay I am going to sleep now probably | 22:58 |
wiking | gnite | 22:58 |
blackburn | I am going to be at job 10 (11) o'clock :D | 22:59 |
blackburn | or sth like 10* | 22:59 |
n4nd0 | blackburn: good night | 23:00 |
-!- blackburn1 [~blackburn@188.168.5.89] has joined #shogun | 23:00 | |
blackburn1 | oops my cable didn't like that | 23:00 |
blackburn1 | wiking: I think git-cc is too dangerous to even try | 23:00 |
blackburn1 | CC looks like a space ship command panel | 23:01 |
wiking | :))) | 23:01 |
blackburn1 | who knows may be one button could destroy the whole world | 23:01 |
wiking | yeah i think it opens a new dimension if u try it :))) | 23:01 |
blackburn1 | one may end with nuclear war or some chemical disaster | 23:01 |
wiking | if saddam would have used gitcc | 23:02 |
n4nd0 | haha you guys scare me sometimes :D | 23:02 |
wiking | maybe they would have found some weapons of mass destructions :D | 23:02 |
blackburn1 | yeah CC with development stream | 23:02 |
blackburn1 | with all files hijacked | 23:02 |
blackburn1 | n4nd0: you should try CC | 23:03 |
blackburn1 | to achieve zen | 23:03 |
blackburn1 | in hofstadter book a theory that zen's stuff is just for destroy of all means | 23:03 |
-!- blackburn [~blackburn@188.168.13.61] has quit [Ping timeout: 265 seconds] | 23:04 | |
blackburn1 | CC is just the same, it destroys development as we know it | 23:04 |
n4nd0 | cc is a version control system right? | 23:04 |
blackburn1 | yes | 23:04 |
wiking | n4nd0: nonono | 23:04 |
wiking | blackburn1: dont confuse him | 23:04 |
blackburn1 | spaceship missile control system | 23:04 |
n4nd0 | I was confused at the beginning with the command cc | 23:04 |
blackburn1 | with version control support | 23:04 |
wiking | n4nd0: it's like a nuclear reactor man | 23:04 |
n4nd0 | hahaha | 23:05 |
blackburn1 | have you seen star trek? | 23:05 |
n4nd0 | no | 23:06 |
blackburn1 | may be gayniggers from the outer space? | 23:06 |
blackburn1 | they had planet destruction stuff | 23:07 |
blackburn1 | it comes out of box with clear case | 23:07 |
n4nd0 | I don't know what the heck are you talking about :) | 23:07 |
wiking | :D | 23:08 |
wiking | n4nd0: never ever touch clearcase man | 23:08 |
wiking | not even in your dreams | 23:08 |
n4nd0 | all right, I promise you that :) | 23:08 |
n4nd0 | I had heard about subversion and cvs but not about clearcase | 23:08 |
n4nd0 | when was that used? | 23:08 |
blackburn1 | n4nd0: in 1400-1450 | 23:09 |
blackburn1 | when the inquisition was like fashion | 23:09 |
wiking | hahahahha | 23:09 |
wiking | blackburn1: columbus tried to teach CC for the indians | 23:09 |
wiking | and they told him: i rather fucking die | 23:09 |
n4nd0 | :DD | 23:09 |
wiking | ok i know i might be a bit offensive with this comment, wasn't inteded | 23:10 |
n4nd0 | then it was probably invented by Spanish people | 23:10 |
blackburn1 | n4nd0: it is not true jeanne d'arc was burnt | 23:10 |
n4nd0 | Spain was quite an important country back then | 23:10 |
blackburn1 | she had to commit to UNOPENABLE DEVELOPMENT STREAM | 23:10 |
n4nd0 | is that like a code in that system? | 23:11 |
wiking | n4nd0: i think they thought that using CC would be the solution... that's why your ship sank against nelson admiral | 23:11 |
blackburn1 | yeah, they just forgot to complete VOBs on integration views | 23:12 |
blackburn1 | that usually leads to ship sink or sth like that | 23:12 |
blackburn1 | chernobyl is the similar case | 23:12 |
wiking | :DDDD | 23:12 |
blackburn1 | somebody joined wrong project | 23:12 |
n4nd0 | haha | 23:12 |
wiking | blackburn1: i bet gorbachov thought that using CC would be the solution as well | 23:12 |
n4nd0 | I think we should move shogun to CC | 23:12 |
blackburn1 | haha lol | 23:12 |
n4nd0 | fuck off with git | 23:13 |
wiking | :D | 23:13 |
n4nd0 | CC sounds exciting | 23:13 |
blackburn1 | it is | 23:13 |
alexlovesdata | bye guys | 23:13 |
-!- alexlovesdata [~binder@goldenezahl.ml.tu-berlin.de] has left #shogun [] | 23:13 | |
wiking | CC is a masterpiece | 23:13 |
blackburn1 | already leaving? | 23:13 |
blackburn1 | ahh | 23:13 |
n4nd0 | we scared alex! | 23:13 |
wiking | i must check images.google for CC!!! | 23:13 |
blackburn1 | okay now finally leaving you too | 23:14 |
n4nd0 | http://www.cmcrossroads.com/forum/31-clearcase/100442-clearcase-ucm-vs-git | 23:14 |
n4nd0 | CC vs git | 23:14 |
blackburn1 | have a nice night I will push you tomorrow to develop more code with clearcase | 23:14 |
blackburn1 | oh | 23:15 |
blackburn1 | lock before edit is a pros | 23:15 |
blackburn1 | is in pros* | 23:15 |
blackburn1 | that's really nice | 23:15 |
blackburn1 | wiking: fits for your issue about breaking code :D | 23:15 |
n4nd0 | First, I would answer that comparing git to UCM is unfair to ClearCase, and not only from the point of view of performance! | 23:16 |
blackburn1 | nobody would break your code because you can LOCK IT | 23:16 |
blackburn1 | :D ohmfg | 23:16 |
wiking | :DDD | 23:16 |
blackburn1 | n4nd0: they all sound like dinosaurs | 23:16 |
blackburn1 | I've seen one BIG COMPANY WITH MANAGERS claimed that open source solutions sucks | 23:17 |
blackburn1 | recently | 23:17 |
blackburn1 | in 2012 | 23:17 |
wiking | blackburn1: yes | 23:17 |
blackburn1 | gosh | 23:17 |
wiking | blackburn1: that's pretty much normal :D | 23:17 |
n4nd0 | haha | 23:17 |
blackburn1 | okay okay | 23:17 |
n4nd0 | I found an answer you would like more in another forum | 23:17 |
blackburn1 | see you tomorrow | 23:17 |
blackburn1 | :) | 23:17 |
n4nd0 | Now, there is some potential in ClearCase that might make it compare favourably to Git. | 23:17 |
wiking | duuuude | 23:17 |
wiking | " If not, I'll probably need to have some internal discussion before proceeding. | 23:17 |
n4nd0 | But it is not commonly known or used, and would require some support at last from IBM. | 23:17 |
wiking | these guys are trying to use | 23:18 |
n4nd0 | I love the last part | 23:18 |
wiking | shogun @ ey | 23:18 |
wiking | :DDD | 23:18 |
blackburn1 | n4nd0: yeah nuclear missile button is absent in git | 23:18 |
blackburn1 | :D | 23:18 |
n4nd0 | https://www.ibm.com/developerworks/forums/thread.jspa?messageID=14787484� | 23:18 |
wiking | blackburn1: :DDDDDDDDDDDDDDDDDDDDD | 23:19 |
wiking | ROTFL | 23:19 |
wiking | :DDDD | 23:19 |
blackburn1 | okay see you :) | 23:19 |
-!- blackburn1 [~blackburn@188.168.5.89] has quit [Quit: Leaving.] | 23:19 | |
wiking | ok | 23:20 |
wiking | let's support EY to use shoung | 23:20 |
wiking | :D | 23:20 |
wiking | erm shogun ;P | 23:20 |
n4nd0 | sure | 23:21 |
n4nd0 | wiking: do you know it gcc 4.1 will go well with shogun v.1.1.0? | 23:22 |
wiking | yeah | 23:23 |
wiking | i'm just checking | 23:23 |
wiking | that why actually we had that bug | 23:23 |
wiking | i must dig out the stupid commit log | 23:23 |
wiking | ok yeah htis one | 23:27 |
wiking | mmmm http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23287 | 23:27 |
wiking | n4nd0: can u open this one? | 23:27 |
n4nd0 | yes | 23:28 |
n4nd0 | it took a bit long though | 23:28 |
wiking | yeah here too | 23:28 |
wiking | mmm what was the code that actually caused this one | 23:28 |
wiking | ah git grep will tell me | 23:28 |
n4nd0 | oh didn't know there is a grep in git | 23:28 |
wiking | yeah | 23:29 |
wiking | it's fucking great | 23:29 |
wiking | ;) | 23:29 |
n4nd0 | sound really useful yes | 23:29 |
n4nd0 | thank you! | 23:29 |
wiking | awesomefast | 23:29 |
wiking | mmm do we have tags? | 23:29 |
wiking | ah yeah cool we do | 23:29 |
n4nd0 | what are you executing exactly? | 23:30 |
wiking | trying to find out | 23:30 |
wiking | whether that bug actually present in 1.1.0 version | 23:30 |
n4nd0 | what command are you using? just to learn more about it | 23:30 |
wiking | well now i've done this | 23:30 |
wiking | git grep "\.~" | 23:31 |
wiking | that'll give me back where the destructor was called explicitly in the code | 23:31 |
wiking | and now i wanna do the same | 23:31 |
wiking | on the 1.1.0 tag | 23:31 |
wiking | but i cannot find tag for that | 23:31 |
n4nd0 | aham I see | 23:31 |
wiking | git tag | 23:31 |
n4nd0 | did you go back to that commit? | 23:31 |
wiking | n4nd0: yeah tag is for that | 23:32 |
n4nd0 | what was the gcc bug exactly? | 23:32 |
wiking | n4nd0: that you cannot call explicitly a destructor | 23:32 |
wiking | on a templated class | 23:32 |
n4nd0 | all right | 23:32 |
wiking | i.e. something.~SGVector<float64_t>() | 23:32 |
wiking | will not compile | 23:32 |
n4nd0 | ok | 23:32 |
wiking | the problem is that there's no tag for 1.1.0 release | 23:33 |
n4nd0 | too bad | 23:33 |
wiking | i'll check 1.0.0 | 23:33 |
n4nd0 | maybe with the release date we can guess the last commit done by that time? | 23:33 |
wiking | ok in 1.0.0 there was no such thing | 23:34 |
wiking | mmm i cannot see it from the git log | 23:35 |
wiking | which one supposed to be the 1.1.0 release | 23:35 |
wiking | ok | 23:36 |
wiking | how the fuck he downloaded 1.1.0 ? | 23:36 |
wiking | ah ok | 23:36 |
wiking | i was checking the wrong thingy | 23:36 |
wiking | mmm | 23:36 |
wiking | i really dont feel for downloading the tar.gz and do a grep -r on it | 23:37 |
wiking | ;) | 23:37 |
wiking | i'll just tell him to try to compile it with 4.1 | 23:37 |
wiking | and it might fail | 23:37 |
wiking | if it does | 23:37 |
wiking | then it's because we execute that bug :D | 23:37 |
n4nd0 | but | 23:37 |
n4nd0 | wasn't the bug fixed in gcc 4? | 23:37 |
wiking | n4nd0: after 4.3 | 23:38 |
n4nd0 | ahh al right | 23:38 |
n4nd0 | I thought it was after 3.x | 23:38 |
wiking | noup | 23:38 |
wiking | 4.3 and later | 23:38 |
wiking | n4nd0: check the comments on the bug page | 23:38 |
wiking | ok this guy might be lucky | 23:39 |
wiking | i've just checked it :D | 23:39 |
n4nd0 | let's see | 23:40 |
wiking | ok | 23:42 |
wiking | support email sent | 23:42 |
n4nd0 | good job, nice mail! | 23:43 |
wiking | hahah another one | 23:44 |
wiking | osx | 23:44 |
wiking | it should fucking work | 23:44 |
wiking | ok shit i forgot to cc the mailinglist | 23:45 |
wiking | ;( | 23:45 |
wiking | wonder what acient osx is he using | 23:46 |
n4nd0 | time to sleep for me | 23:47 |
n4nd0 | good night! | 23:47 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 23:47 | |
--- Log closed Thu Aug 23 00:00:17 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!