--- Log opened Wed May 02 00:00:37 2012 | ||
CIA-64 | shogun: Sergey Lisitsyn master * r2c8ae73 / (4 files in 2 dirs): Proper retraining of multiclass - http://git.io/RYQGsg | 00:50 |
---|---|---|
CIA-64 | shogun: Sergey Lisitsyn master * r800be9b / (src/shogun/regression/LARS.cpp src/shogun/regression/LARS.h): | 00:50 |
CIA-64 | shogun: Revert "Removed unnecessary methods from LARS" | 00:50 |
CIA-64 | shogun: This reverts commit ca57115f7cad3b447fe4125299039ef17291ffb2. - http://git.io/ptz6yA | 00:50 |
-!- Netsplit *.net <-> *.split quits: @sonney2k, wiking, naywhayare | 00:56 | |
-!- blackburn [~qdrgsm@188.168.4.67] has quit [Ping timeout: 260 seconds] | 00:58 | |
-!- gsomix [~gsomix@188.168.13.108] has quit [Quit: Ex-Chat] | 00:59 | |
-!- Netsplit over, joins: wiking, naywhayare | 01:00 | |
-!- sonney2k [~shogun@7nn.de] has joined #shogun | 01:02 | |
-!- ServerMode/#shogun [+o sonney2k] by hitchcock.freenode.net | 01:02 | |
-!- CIA-113 [~CIA@cia.atheme.org] has joined #shogun | 01:48 | |
-!- CIA-64 [~CIA@cia.atheme.org] has quit [Ping timeout: 272 seconds] | 01:50 | |
-!- av3ngr [av3ngr@nat/redhat/x-ygbjirmhcsmqgnqq] has joined #shogun | 02:40 | |
-!- vikram360 [~vikram360@117.192.177.84] has joined #shogun | 04:41 | |
-!- vikram360 [~vikram360@117.192.177.84] has quit [Ping timeout: 272 seconds] | 06:48 | |
-!- vikram360 [~vikram360@117.192.176.148] has joined #shogun | 06:53 | |
-!- vikram360 [~vikram360@117.192.176.148] has quit [Ping timeout: 246 seconds] | 06:58 | |
-!- vikram360 [~vikram360@117.192.183.45] has joined #shogun | 06:59 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 07:53 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 07:56 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 07:56 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 07:56 | |
-!- sonne|work [~sonnenbu@194.78.35.195] has joined #shogun | 08:18 | |
-!- av3ngr [av3ngr@nat/redhat/x-ygbjirmhcsmqgnqq] has quit [Quit: That's all folks!] | 09:19 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 09:42 | |
-!- av3ngr [~av3ngr@58.108.72.131] has joined #shogun | 09:51 | |
-!- av3ngr [~av3ngr@58.108.72.131] has quit [Read error: Connection reset by peer] | 10:40 | |
-!- pluskid [~pluskid@111.120.7.70] has joined #shogun | 11:05 | |
sonne|work | hey pluskid! | 11:05 |
sonne|work | thanks for your comments about sgvector | 11:05 |
pluskid | sonne|work: so is the new sgvector available now? | 11:06 |
sonne|work | we will do the interface variant first and then later do the rename SGIVector / SGVector if we have some good usecases | 11:06 |
pluskid | cool | 11:06 |
sonne|work | pluskid: no not yet | 11:06 |
sonne|work | but soon | 11:06 |
pluskid | hope to see the new classes, because it is really used everywhere :D | 11:07 |
sonne|work | pluskid: the other big transition I think we should do rather soon is the CLabel split up | 11:07 |
sonne|work | pluskid: it will break things... | 11:07 |
sonne|work | (sgvector I mean) | 11:07 |
pluskid | yes, we got some discussions, but stops then | 11:07 |
sonne|work | we will probably spend a week or so fixing leaks / etc | 11:08 |
pluskid | sonne|work: so we hope it breaks ASAP, so we can fix those soon :) | 11:08 |
sonne|work | exactly | 11:08 |
sonne|work | such that when gsoc starts we are back to stable again | 11:08 |
sonne|work | pluskid: so regarding labels I think an easy start is to derive multiclass and RealLabels and Binary Labels from current labels | 11:09 |
pluskid | yes, the memory usage will be very difficult to track because it is rather unpredictable of where the memory is allocated/freed in the previous SGVector | 11:09 |
sonne|work | yes :( | 11:09 |
sonne|work | I have some ideas but sometimes it is DKJ$%H$OH%!! | 11:09 |
sonne|work | pluskid: would you want to do the CLabels change? | 11:09 |
sonne|work | not sure if I am asking to much here... | 11:09 |
pluskid | sonne|work: sure | 11:10 |
sonne|work | but all these hacks with checking if label is 2 class | 11:10 |
pluskid | sonne|work: did you know the draft design we discussed about the other day? | 11:10 |
sonne|work | binary or so suck | 11:10 |
sonne|work | yes | 11:10 |
sonne|work | hence the RealLabels name :) | 11:10 |
pluskid | yes | 11:10 |
sonne|work | pluskid: what I think is that the current clabels class will work as-is for binary, mc, and regression | 11:10 |
pluskid | yes | 11:11 |
sonne|work | so maybe all that is needed are some overloaded methods that check whether the set label vector is valid | 11:11 |
pluskid | you mean something like is_valid_twoclass ? | 11:11 |
sonne|work | I don't (currently) think we need some extra IntLabels - internally things could still be doubles | 11:12 |
sonne|work | yeah | 11:12 |
sonne|work | but using a more generic name | 11:12 |
sonne|work | like is_valid() | 11:12 |
sonne|work | and that check should be done when label vector is set / created | 11:12 |
sonne|work | so constructor and set function only | 11:12 |
pluskid | you mean, for example, in MulticlassMachine, is_valid ensures the labels are at least 3 classes? | 11:13 |
sonne|work | no | 11:13 |
sonne|work | multiclassmachine uses CMulticlassLabels | 11:13 |
pluskid | oh, I see, you mean we still store and set/get as float | 11:13 |
sonne|work | and is_valid would only check if the set labels are integers >= 0 | 11:13 |
pluskid | hmm | 11:14 |
pluskid | I sse | 11:14 |
pluskid | see | 11:14 |
sonne|work | mc labels would need extra functions like get_num_classes etc | 11:14 |
pluskid | but maybe someone needs to extend more, like in SO learning? | 11:14 |
pluskid | then a float might not be enough? | 11:15 |
sonne|work | exactly | 11:15 |
sonne|work | but it is totally unclear how that would look like | 11:15 |
sonne|work | i.e., which shared functions exist at all? | 11:15 |
pluskid | yes | 11:15 |
pluskid | actually, I'm wondering do they have to have a shared baseclass? | 11:16 |
sonne|work | yeah only for the look of it I guess :) | 11:16 |
sonne|work | maybe is_valid would still make sense | 11:16 |
sonne|work | and maybe even some way to check if a predicted label == true label | 11:17 |
sonne|work | so some way to compare a label | 11:17 |
pluskid | yes | 11:17 |
pluskid | that is important for SO | 11:17 |
pluskid | and some way to compute distance | 11:17 |
pluskid | of labels | 11:17 |
sonne|work | though there are tons of ways | 11:17 |
sonne|work | and they can be arbitrarily complex | 11:17 |
pluskid | the problem is, if the CLabels do not have a shared base class | 11:18 |
sonne|work | pluskid: I would say we start with CLabels -> MC,RealLabels,Binary | 11:18 |
pluskid | we might want to move methods like apply() set_labels() in CMachine to proper subclasses like CMulticlassMachine, CSOMachine, etc. | 11:19 |
sonne|work | and if we realize we can share things we could do CLabels -> SOLabels, (RealLabels -> MC, Binary) | 11:19 |
pluskid | maybe have to create a CBinaryMachine and CRegressionMachine | 11:19 |
sonne|work | makes sense - I am worried that this will explode our class hierachy though | 11:20 |
sonne|work | it is certainly more clean... | 11:20 |
pluskid | yeah, CMachine becomes more abstract | 11:20 |
pluskid | in this way, we no longer have to worry about a common base class of different kind of labels | 11:21 |
pluskid | CMulticlassMachine will use CMulticlassLabels | 11:21 |
pluskid | CBinaryClassMachine will use CBinaryLabels | 11:21 |
sonne|work | we had the same problem with CMachine (in apply) using CFeatures | 11:22 |
sonne|work | but there we just just CFeatures and check at runtime that things match | 11:22 |
pluskid | hmm, but Features are much more complicated than labels I think? | 11:22 |
sonne|work | pluskid: what i mean is that machines usually work with one type of features | 11:22 |
sonne|work | some machines (kernel machines) can work with arbitrary types | 11:23 |
pluskid | sonne|work: hmm, I see | 11:23 |
pluskid | sonne|work: but particular kind of machine *only* work with particular kind of labels | 11:23 |
sonne|work | pluskid: anyway I think the discussion about split up of machine by label can be postponed | 11:23 |
sonne|work | yeah maybe | 11:24 |
pluskid | I'll write an email to the mailing list | 11:24 |
sonne|work | but think of some machine | 11:24 |
sonne|work | for which you can set the loss function | 11:24 |
sonne|work | that one will work also for different labels | 11:24 |
sonne|work | hah | 11:24 |
sonne|work | actually the new liblinear would be an example | 11:24 |
pluskid | hmm | 11:24 |
sonne|work | one can do regression or binary or mc classification with it | 11:25 |
sonne|work | of course one can split things up and make them explicit | 11:25 |
sonne|work | and we did mostly so far | 11:25 |
pluskid | but we have 3 wrappers for liblinear for regression, mc or binary, don't we? | 11:25 |
pluskid | you mean in the future we might wrap LibLinear as a generic CMachine? | 11:26 |
sonne|work | 2 yes | 11:26 |
sonne|work | I like the split | 11:26 |
sonne|work | it makes things more explicit | 11:26 |
sonne|work | but in future we might have some machine that you just give a CLoss object | 11:27 |
pluskid | hmm, thanks, I'll try to summarize and write an email to the list to ask other people's opinion | 11:27 |
sonne|work | and then it can do anything | 11:27 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 11:27 | |
sonne|work | so I am not in favor of splitting Machines by label | 11:27 |
blackburn | oh some discussion there | 11:27 |
blackburn | enlighten me ;) | 11:27 |
sonne|work | but clabels only | 11:27 |
pluskid | sonne|work: it's currently OK with reg, binary and mc | 11:28 |
sonne|work | blackburn: I asked pluskid if he could do the other big transition (clabel splitup) | 11:28 |
pluskid | but if we like to add SO | 11:28 |
blackburn | sonne|work: btw we need to help n4ndo with it | 11:28 |
blackburn | he is a little stucked with that design decision | 11:28 |
blackburn | how can SO labels look like? | 11:28 |
pluskid | currently, we have get_label(int), and get_int_label(int), if we have SO, we might need get_AAA_label(int), get_BBB_label(int) something like those in CLabels | 11:29 |
sonne|work | pluskid: then we add them in there - what is the problem? | 11:29 |
blackburn | I am ok to add some method there | 11:29 |
sonne|work | pluskid: we have to cast labels | 11:29 |
blackburn | get_structured_label | 11:29 |
sonne|work | anyways | 11:29 |
blackburn | yeah no way w/o casting probably | 11:29 |
pluskid | blackburn: what it return? a CSGObject ? | 11:30 |
blackburn | pluskid: yes, or some StructuredOutput base class | 11:30 |
sonne|work | can be anything | 11:30 |
sonne|work | graph | 11:30 |
pluskid | sonne|work: ok, I see | 11:30 |
sonne|work | segmentation | 11:30 |
pluskid | so dynamic_cast again | 11:30 |
pluskid | like with features | 11:30 |
sonne|work | (so vector/matrix could suffice ...) | 11:30 |
sonne|work | pluskid: well with CFeatures we explicitly check for type | 11:31 |
pluskid | then why would we bother splitting? | 11:31 |
sonne|work | and then cast | 11:31 |
blackburn | pluskid: I had one concern about lars | 11:31 |
sonne|work | pluskid: well there might be some shared functions still | 11:31 |
pluskid | sonne|work: dynamic_cast will return NULL if it is not the subclass type you want | 11:31 |
blackburn | let us discuss it after that issue | 11:31 |
sonne|work | like is_valid | 11:31 |
blackburn | get_number of labels is shared at leasy | 11:31 |
blackburn | least* | 11:31 |
pluskid | hmm | 11:32 |
pluskid | becomes more clear now | 11:32 |
sonne|work | pluskid: I know but one cannot check which type an object has in C++ in a compiler independent way | 11:32 |
pluskid | sonne|work: hmm, that's true | 11:32 |
sonne|work | so it is better to put an enum in there | 11:32 |
pluskid | leave for supper | 11:33 |
sonne|work | enjoy! | 11:33 |
pluskid | be back in a while | 11:33 |
blackburn | sonne|work: I have one idea about backcompatibility | 11:33 |
blackburn | Labels should be a factory | 11:33 |
blackburn | Labels([1,2,3]) => MulticlassLabels | 11:33 |
sonne|work | uii | 11:33 |
blackburn | Labels([graph1,graph2]) => SO labels | 11:33 |
sonne|work | intelligence in labels :D | 11:34 |
blackburn | no, hardcode rather | 11:34 |
blackburn | but compatibility would be here | 11:34 |
blackburn | code compatibility | 11:34 |
blackburn | as for serialization.. | 11:34 |
blackburn | I'd rather let users convert their object by theyselves.. | 11:35 |
blackburn | objects* | 11:35 |
blackburn | it becomes too difficult to handle | 11:35 |
sonne|work | yeah true serialization will be b0rked | 11:38 |
pluskid | but we cannot produce object of another class from a constructor | 11:52 |
pluskid | btw, in the data/toy directory | 11:54 |
pluskid | fm_train_real.dat label_train_multiclass.dat are data and label for training | 11:54 |
pluskid | fm_test_real.dat is data for testing | 11:55 |
pluskid | but do we have labels for testing? | 11:55 |
-!- karlnapf [~heiko@host86-179-192-128.range86-179.btcentralplus.com] has joined #shogun | 11:55 | |
blackburn | pluskid: no, I don't mean C++ class for that | 11:55 |
blackburn | probably python function Labels() would work too | 11:56 |
blackburn | I know how to do that at least in java and python | 11:56 |
pluskid | in Java you can new CLabel() and get a CMulticlassLabel ? | 11:57 |
blackburn | pluskid: about LARS - don't you think it should be LeastAnglesRegression? | 11:57 |
pluskid | blackburn: you mean the name? | 11:57 |
pluskid | I'm OK with both | 11:57 |
blackburn | pluskid: yeah | 11:57 |
blackburn | I think we should prefer *Regression naming | 11:57 |
blackburn | I do work on some regression right now | 11:57 |
blackburn | FeatureTreeLeastSquaresRegression namely | 11:58 |
pluskid | haha, our class names are becoming longer :) | 11:58 |
blackburn | pluskid: I have spent a lot of time but no idea how to do that shorter.. | 11:58 |
blackburn | however in my multitask part this naming could be convenient | 11:59 |
blackburn | next class to implement is TaskTreeLeastSquaresRegression | 11:59 |
blackburn | pluskid: so I have one ask - could you please rename it and provide constructor taking (CFeatures*,CLabels*) | 12:00 |
blackburn | ? | 12:00 |
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun | 12:00 | |
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host] | 12:00 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 12:00 | |
pluskid | blackburn: don't worry, at least our names are shorter than org.aspectj.weaver.patterns.HasThisTypePatternTriedToSneakInSomeGenericOrParameterizedTypePatternMatchingStuffAnywhereVisitor | 12:00 |
blackburn | pluskid: lol | 12:00 |
pluskid | blackburn: LARS? | 12:00 |
blackburn | pluskid: yeap | 12:01 |
pluskid | OK | 12:01 |
blackburn | pluskid: and also please remove save/load | 12:01 |
pluskid | that's cool | 12:01 |
pluskid | I hate them :D | 12:01 |
blackburn | and get_w that does store to given pointer | 12:01 |
blackburn | only SGVector should stay | 12:01 |
blackburn | oh I hate shindows :E | 12:02 |
blackburn | 2 minutes to force reboot | 12:02 |
pluskid | blackburn: wait a minute, I forgot about the code | 12:02 |
blackburn | it says it will reboot automatically in 1.23 | 12:02 |
blackburn | :D | 12:02 |
blackburn | and restart later is inactive | 12:02 |
pluskid | ... | 12:02 |
pluskid | you mean what get_w? | 12:03 |
blackburn | pluskid: float64_t*& | 12:03 |
blackburn | 0:05 | 12:03 |
blackburn | haha | 12:03 |
pluskid | ok, see you later | 12:03 |
blackburn | nooooo | 12:03 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Ping timeout: 245 seconds] | 12:08 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 12:15 | |
blackburn | pluskid: re | 12:15 |
pluskid | blackburn: where's regression_lars.py in examples? | 12:16 |
pluskid | is it removed? | 12:16 |
blackburn | hmm I didn't move or remove it | 12:17 |
pluskid | Oh, I found it | 12:17 |
pluskid | it is moved to graphical folder | 12:17 |
pluskid | btw: what do you mean by get_w? | 12:17 |
pluskid | I wrote that in the way CLinearMachine implemented | 12:18 |
blackburn | pluskid: why? | 12:18 |
blackburn | because of nz parameter? | 12:18 |
blackburn | pluskid: https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/regression/LARS.h | 12:18 |
blackburn | line 157 | 12:18 |
blackburn | this method is not needed (I know it is in linearmachine but should be removed as well) | 12:19 |
pluskid | so you want to copy into the pointer, instead of simply assigning? | 12:20 |
pluskid | OK | 12:20 |
pluskid | you mean remove it? | 12:20 |
pluskid | I see | 12:20 |
blackburn | pluskid: yes no need to have it because it won't work in modular | 12:21 |
pluskid | <pluskid> btw, in the data/toy directory | 12:25 |
pluskid | <pluskid> fm_train_real.dat label_train_multiclass.dat are data and label for training | 12:25 |
pluskid | <pluskid> fm_test_real.dat is data for testing | 12:25 |
pluskid | <pluskid> but do we have labels for testing? | 12:25 |
pluskid | blackburn: do you know this? ^^ | 12:25 |
blackburn | pluskid: yes, no testing labels | 12:26 |
pluskid | hmm, strange... | 12:26 |
pluskid | then no way to compute the prediction error | 12:26 |
blackburn | pluskid: we rather compare results are the same | 12:28 |
pluskid | blackburn: I just want to get a sense of how various ECOC scheme performs :) | 12:28 |
blackburn | hmm I can give you some data if you want | 12:30 |
pluskid | blackburn: is there any pre-existing in the data directory? | 12:30 |
pluskid | that submodule is quite large | 12:30 |
blackburn | no, our data is rather toy | 12:30 |
blackburn | there are some applications data and testing data | 12:30 |
blackburn | but I do not give a f how to operate on genome data | 12:30 |
pluskid | give me some link to mldata or UCI is also OK | 12:31 |
pluskid | I don't want big data | 12:31 |
blackburn | why? | 12:31 |
pluskid | just some local play | 12:31 |
pluskid | on my laptop | 12:31 |
pluskid | to get some basic sense | 12:31 |
pluskid | instead of to publish a paper :p | 12:32 |
pluskid | I wish I can run the experiment within several seconds | 12:32 |
blackburn | heh you can't really compare accuracy on small dataset I think | 12:33 |
pluskid | hmm | 12:34 |
pluskid | then large data is also OK | 12:34 |
pluskid | as long as my laptop can run it :) | 12:34 |
sonne|work | blackburn: sure you can :) | 12:35 |
blackburn | sonne|work: ECOC vs OvR? lets see.. | 12:35 |
blackburn | pluskid: I will compare your ecoc on my data soon | 12:36 |
blackburn | would be clear if it works correct and even if it gives any extra performance | 12:36 |
sonne|work | why not use mnist data? | 12:36 |
sonne|work | (some subset is totally enough) | 12:36 |
sonne|work | pluskid: looks like we are too late for winning http://lshtc.iit.demokritos.gr/ | 12:37 |
sonne|work | but it also has some nice mc data with *lots* of classes | 12:38 |
pluskid | :D | 12:38 |
blackburn | pluskid: that mnist suggestion of sonne|work is pretty good, it is more 'standard' | 12:38 |
pluskid | blackburn: I see, I have that data | 12:39 |
blackburn | pluskid: but I'll let you know how does it work on my HOG data | 12:39 |
blackburn | sonne|work: and you ;) | 12:39 |
pluskid | I usually use the raw data | 12:40 |
pluskid | maybe feature extraction will produce better performance | 12:41 |
blackburn | sonne|work: I still think ECOC would hardly work on small data | 12:43 |
pluskid | blackburn: here comes the LARS renaming pull request | 12:46 |
pluskid | sonne|work: I'll work on CLabels refactoring once I added some python_modular examples for Sparse/Dense Random ECOC | 12:46 |
pluskid | work progress would be a bit slow these days because I'm with a serious cough | 12:47 |
pluskid | hope to cure it soon :) | 12:47 |
blackburn | pluskid: thanks | 12:47 |
blackburn | pluskid: oh get well soon | 12:47 |
blackburn | you already did great job on ecoc | 12:47 |
pluskid | blackburn: maybe we already have something similar to that of scikit, but I haven't look at the data-dependent ECOC coding in ECOClib yet | 12:48 |
pluskid | that would be a big work I guess | 12:48 |
blackburn | pluskid:it looks like all you would need is to provide more distances | 12:49 |
blackburn | no? | 12:49 |
blackburn | and more coders sure | 12:49 |
pluskid | blackburn: distances are decoders | 12:50 |
pluskid | and there are some data-dependent *en*-coders | 12:50 |
blackburn | pluskid: I see | 12:51 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 265 seconds] | 12:55 | |
sonne|work | pluskid: sad to hear! get well soon | 13:07 |
pluskid | sonne|work: thanks, I hope, too | 13:07 |
sonne|work | blackburn: why not? if you have the right ECOC strategy or even some prior knowledge about the hierarchy things can work better with fewer data | 13:09 |
blackburn | sonne|work: just some gut feeling that can be wrong for sure :) | 13:09 |
blackburn | booring | 13:14 |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 13:22 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 13:23 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 13:23 | |
sonne|work | the mighty wiking...! | 13:25 |
wiking | yeeey | 13:25 |
wiking | i've made it | 13:25 |
wiking | ! | 13:25 |
wiking | !! | 13:25 |
sonne|work | hardcore connecting yeah | 13:26 |
blackburn | sonne|work: I guess he mean his paper ;) | 13:27 |
sonne|work | wiking: congrats to that one too :D | 13:27 |
wiking | sonne|work: hehe yeah i was rather meaning my paper for the conf :P | 13:28 |
wiking | cheers | 13:28 |
wiking | the only thing i have to do now is to find my supervisor so that he can sign the copyright form for the paper since his name is going to be on it, although he did not give a damn about it... | 13:29 |
wiking | isn't it a bit ironic? :) | 13:29 |
blackburn | lool | 13:30 |
blackburn | you are kind of extreme guy | 13:30 |
wiking | and that's even more funny that i cannot reach him, neither phone nor email nothing :D | 13:32 |
wiking | mmm | 14:16 |
wiking | n4nd0 is still in spain? | 14:16 |
blackburn | wiking: yes | 14:17 |
wiking | mm has he told when is he going to be around? | 14:17 |
-!- pluskid [~pluskid@111.120.7.70] has quit [Ping timeout: 250 seconds] | 14:18 | |
-!- pluskid [~pluskid@111.120.7.70] has joined #shogun | 14:18 | |
blackburn | wiking: he is going to have exam on 7th or 8th | 14:19 |
wiking | ah ok | 14:19 |
wiking | so no n4nd0 till the 9th | 14:19 |
wiking | roger that | 14:19 |
blackburn | wiking: did you plan to work on something? | 14:19 |
wiking | yep | 14:20 |
wiking | need to talk with him | 14:20 |
wiking | about SO framework | 14:20 |
wiking | with PSI representation | 14:20 |
blackburn | yeah we need to discuss design there | 14:21 |
blackburn | feel free to send email to him :) | 14:21 |
blackburn | you may also CC me and soeren | 14:21 |
wiking | mmm | 14:23 |
wiking | well i think maybe the mailing list would be the best option, or? | 14:23 |
blackburn | wiking: yes sure | 14:24 |
-!- pluskid [~pluskid@111.120.7.70] has quit [Ping timeout: 276 seconds] | 14:26 | |
-!- pluskid [~pluskid@li400-235.members.linode.com] has joined #shogun | 14:26 | |
sonne|work | wiking: still CCing the relevant people :D | 14:33 |
wiking | sonne|work: ookeey | 14:34 |
-!- eric___ [2e1fd566@gateway/web/freenode/ip.46.31.213.102] has joined #shogun | 14:53 | |
pluskid | aaaaaaaaaaaaah, sick... >_<~~~ | 14:56 |
sonne|work | pluskid: painkiller to the rescue! | 14:59 |
sonne|work | ..and sleep | 15:00 |
eric___ | hi | 15:05 |
eric___ | wiking: hi. comin back from holidays in the Alpes. We talked together last friday on running latent svm on action reco databases, remember ? | 15:06 |
wiking | eric___: yoo yo yes | 15:07 |
wiking | i hope u had fun there ;) | 15:07 |
eric___ | last ski weekend of the year | 15:07 |
eric___ | =) | 15:07 |
eric___ | Last friday I asked questions about my problem of releasing mem with shogun like UNREF(svm) when I initialized SimpleFeature with pointer to opencv data strucure. | 15:09 |
eric___ | with SG_FREE(svm) the problem doesnt occurs, what the difference btw UNREF and FREE ? | 15:09 |
wiking | well unref is for the reference counter | 15:15 |
wiking | free just frees the object | 15:15 |
wiking | but afaik it should be essentially the same | 15:15 |
wiking | as i hope free does consider the reference counter | 15:15 |
wiking | and does not free the object if it's referenced | 15:16 |
-!- pluskid [~pluskid@li400-235.members.linode.com] has quit [Quit: Leaving] | 15:16 | |
sonne|work | eric___: use SG_UNREF for objects derived from SGObject | 15:24 |
eric___ | wiking, sonne|work: thx | 15:28 |
sonne|work | np | 15:30 |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has joined #shogun | 15:36 | |
-!- gsomix [~gsomix@188.168.5.30] has joined #shogun | 15:47 | |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has quit [Ping timeout: 245 seconds] | 15:55 | |
gsomix | hi all | 15:56 |
sonne|work | hi gsomix | 15:56 |
sonne|work | how is your computer doing? | 15:56 |
gsomix | sonne|work, fine. everything is working | 15:57 |
blackburn | sonne|work: could you please your blog post too? :D | 15:58 |
blackburn | fix* | 15:58 |
blackburn | oh main word is missed lol | 15:58 |
sonne|work | !? | 15:59 |
sonne|work | ahh | 15:59 |
sonne|work | name | 15:59 |
sonne|work | gsomix: very good | 15:59 |
blackburn | sonne|work: yeah same issue | 15:59 |
blackburn | thanks | 15:59 |
sonne|work | gsomix: so how about the hashset issue? | 15:59 |
sonne|work | would that be a good start for today? | 16:00 |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has joined #shogun | 16:02 | |
gsomix | sonne|work, I like it. but there is one specific detail in realization - size of table (array of lists for resolving collisions) is const. | 16:03 |
sonne|work | gsomix: you could use a dynarray to grow it if that is needed... | 16:04 |
gsomix | IIRC CSet is used for mallocs counting, right? | 16:04 |
sonne|work | gsomix: currently yes - but when we have a faster hashset other use cases will pop up | 16:05 |
blackburn | sonne|work: why not to use stl::vector? | 16:05 |
blackburn | ah or set | 16:05 |
gsomix | sonne|work, could use a dynarray to grow it | it's not needed (in theory), because I use lists for resolving collisions. maybe I should use another type of resolving. | 16:07 |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: wiking] | 16:10 | |
gsomix | sonne|work, I should think about it a little. because I have written hashmap (hashset) for blackburn, not for common use. | 16:12 |
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun | 16:18 | |
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host] | 16:18 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 16:18 | |
sonne|work | gsomix: just use some fixed hash size (64k or so per default configurable in set constructor) | 16:18 |
sonne|work | gsomix: that will work totally ok | 16:18 |
gsomix | sonne|work, ok. | 16:19 |
blackburn | sonne|work: gsomix: I can hardly remember why I needed it | 16:20 |
gsomix | blackburn, lol :] | 16:21 |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has quit [Ping timeout: 245 seconds] | 16:25 | |
-!- sonne|work [~sonnenbu@194.78.35.195] has quit [Quit: Leaving.] | 16:35 | |
gsomix | blackburn, what about titled features? Can you merge it? | 16:44 |
blackburn | gsomix: yes not now | 16:44 |
blackburn | shit exam tomorrow | 16:45 |
blackburn | I should have prepared but I have not | 16:45 |
gsomix | blackburn, which? | 16:45 |
blackburn | politology | 16:45 |
@sonney2k | blackburn, lenin stalin and all that :) | 17:04 |
blackburn | sonney2k: marx | 17:05 |
CIA-113 | shogun: Chiyuan Zhang master * r9935912 / (7 files in 3 dirs): Rename LARS to LeastAngleRegression - http://git.io/1MfXHQ | 17:05 |
CIA-113 | shogun: Soeren Sonnenburg master * rd587b61 / (7 files in 3 dirs): | 17:05 |
CIA-113 | shogun: Merge pull request #492 from pluskid/rename-LARS | 17:05 |
CIA-113 | shogun: Rename LARS to LeastAngleRegression - http://git.io/GPl_pg | 17:05 |
blackburn | thanks | 17:05 |
blackburn | sonney2k:khruschev destalinized it a little so no more stalin :D | 17:06 |
@sonney2k | gsomix, I dont' understand the feature title patch | 17:07 |
@sonney2k | it removes hashset? | 17:07 |
@sonney2k | introduces hashmap | 17:07 |
@sonney2k | but then uses stl::map!??! | 17:07 |
@sonney2k | did I miss anything? | 17:07 |
blackburn | hahah | 17:07 |
@sonney2k | blackburn, no way | 17:08 |
@sonney2k | stl - why over sudden? | 17:08 |
blackburn | sonney2k: I don't mind stl map there | 17:08 |
blackburn | do you? | 17:09 |
@sonney2k | yeah - you know... | 17:10 |
blackburn | sonney2k: give me a reason :) | 17:10 |
-!- karlnapf [~heiko@host86-179-192-128.range86-179.btcentralplus.com] has quit [Ping timeout: 276 seconds] | 17:11 | |
@sonney2k | what? the remainders of my authority won't help!? | 17:12 |
gsomix | sonney2k, ah. I prepared hashmap, but wondered about the possibility of using stl::map. | 17:12 |
blackburn | sonney2k: heh your authority is there still but reason? | 17:14 |
@sonney2k | blackburn, how do you serialize it? | 17:15 |
@sonney2k | gsomix, I would prefer your hashmap :) | 17:16 |
blackburn | sonney2k: how do you serialize hashmap? | 17:16 |
@sonney2k | if it uses standard data types like SG* stuff it will work | 17:18 |
blackburn | sonney2k: I don't think it is much harder to serialize stl :) | 17:19 |
@sonney2k | it is | 17:19 |
@sonney2k | argh | 17:19 |
@sonney2k | have to leave train | 17:19 |
@sonney2k | cu | 17:19 |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 17:19 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 246 seconds] | 17:25 | |
gsomix | sonney2k, okay master. :) | 17:27 |
gsomix | sonney2k, I need to go. Dance classes. | 17:39 |
gsomix | cu evening | 17:39 |
-!- wiking [~wiking@we02c096.ugent.be] has joined #shogun | 17:40 | |
-!- wiking [~wiking@we02c096.ugent.be] has quit [Changing host] | 17:40 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 17:40 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Ping timeout: 245 seconds] | 17:44 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has joined #shogun | 17:48 | |
-!- wiking [~wiking@78-23-189-112.access.telenet.be] has quit [Changing host] | 17:48 | |
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun | 17:48 | |
wiking | doh | 17:58 |
wiking | wtf is with this latex template | 17:58 |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has joined #shogun | 18:29 | |
eric___ | wiking: simple advice : should I work with double or float is enough precision (for svmmulticlass ..) | 18:53 |
n4nd0 | eric___: hey! I don't know about your particular application but most of the things in Shogun use internally float64_t (aka double) | 18:58 |
n4nd0 | eric___: so if you are trying to avoid some overhead using float instead of double, take that into account | 18:59 |
eric___ | n4nd0: allright, because my training data comes from opencv apps, and I use float32 in it | 19:00 |
eric___ | n4nd0: I should then rewrite the extracting data part with double type .. | 19:01 |
n4nd0 | eric___: if you think that you don't need so high precision, the one offered by float64_t | 19:02 |
n4nd0 | I would just use it as it is | 19:02 |
n4nd0 | it would have a bit more of overhead since float64_t would be used internally in Shogun, but it is probably unimportant | 19:03 |
-!- eric___ [2e1fd566@gateway/web/freenode/ip.46.31.213.102] has quit [Quit: Page closed] | 19:24 | |
-!- blackburn [~qdrgsm@188.168.4.67] has joined #shogun | 19:58 | |
blackburn | hey n4nd0 | 20:02 |
n4nd0 | blackburn: hey! | 20:03 |
blackburn | how are you? | 20:03 |
n4nd0 | I am fine | 20:04 |
n4nd0 | tired of wireless communications :P | 20:04 |
n4nd0 | what about you? | 20:04 |
blackburn | n4nd0: next exam? | 20:05 |
blackburn | I have one tomorrow | 20:05 |
n4nd0 | blackburn: yeah, next exam on that | 20:05 |
n4nd0 | blackburn: what is the topic? | 20:05 |
blackburn | n4nd0: politology :D | 20:05 |
n4nd0 | wtf | 20:05 |
n4nd0 | what is it? | 20:06 |
blackburn | kind of social science I guess | 20:06 |
n4nd0 | oh like politics | 20:06 |
blackburn | yeap | 20:06 |
blackburn | studying politics | 20:06 |
n4nd0 | I was thinking of something related to math | 20:06 |
n4nd0 | :P | 20:06 |
blackburn | haha | 20:06 |
blackburn | I'd rather listen to tom waits | 20:09 |
blackburn | I do not want to study anything :D | 20:09 |
n4nd0 | yeah, I understand | 20:10 |
blackburn | n4nd0: when will you get past exams results? | 20:11 |
n4nd0 | no idea | 20:13 |
blackburn | strange | 20:13 |
blackburn | we have it immediate | 20:13 |
n4nd0 | at most two weeks after I did them | 20:13 |
n4nd0 | so probably I will get to know about them by next week | 20:14 |
* gsomix have returned from dance floor. | 20:14 | |
blackburn | n4nd0: two? wow | 20:14 |
n4nd0 | blackburn: yeah, it is about the same in Sweden normally | 20:14 |
n4nd0 | I think that at the moment is superfast! | 20:15 |
blackburn | n4nd0: heh students here would die of heart attack if professor said he will let they know about results in two weeks | 20:15 |
n4nd0 | haha | 20:16 |
gsomix | blackburn, where are you? | 20:16 |
blackburn | gsomix: 315, need to discuss anything? I was going to start preparing for my exam | 20:16 |
gsomix | blackburn, cake. just cake. :] | 20:17 |
blackburn | cake? | 20:17 |
gsomix | yep | 20:17 |
blackburn | heh | 20:17 |
gsomix | tasty cake | 20:17 |
blackburn | I am not really hungry but I'll visit you if I am ;) | 20:18 |
gsomix | blackburn, ok. :] | 20:18 |
n4nd0 | do you like to cook cakes gsomix? | 20:19 |
n4nd0 | I think it is not the first time I see you guys are talking about it ;) | 20:19 |
blackburn | he get one from home I guess haha | 20:19 |
gsomix | n4nd0, nope. I do not know how. | 20:19 |
gsomix | n4nd0, I just have one. :) | 20:20 |
blackburn | he steal it from the shop | 20:20 |
blackburn | :D | 20:20 |
n4nd0 | :) | 20:20 |
@sonney2k | blackburn, btw - shouldn't we get rid of the load/save functions? | 20:20 |
blackburn | sonney2k: yes we should | 20:21 |
@sonney2k | I mean the serialization stuff should be enough | 20:21 |
blackburn | sonney2k: is it used in static? | 20:21 |
@sonney2k | only problem is we need some replacement for it in static | 20:21 |
@sonney2k | hehe | 20:21 |
blackburn | hmmm | 20:21 |
@sonney2k | there is no way to load anything from cmdline | 20:21 |
@sonney2k | woudln't be too much effort | 20:22 |
@sonney2k | just serialize the single trained model that is there | 20:22 |
blackburn | sonney2k: you have slave for that :D | 20:22 |
@sonney2k | haha | 20:22 |
@sonney2k | gsomix, some easy job - just remove all the load/save functions from Machine and derived classes | 20:23 |
gsomix | n4nd0, I know how to cook pelmeni. ;] | 20:23 |
gsomix | sonney2k, ok. | 20:23 |
* sonney2k loves pelmeni | 20:23 | |
blackburn | sonney2k: really? | 20:23 |
@sonney2k | gsomix, please send some over | 20:23 |
gsomix | sonney2k, huh | 20:24 |
gsomix | rabotatch! | 20:24 |
n4nd0 | gsomix: no idea what it is, let me check | 20:25 |
gsomix | n4nd0, http://en.wikipedia.org/wiki/Pelmeni | 20:25 |
@sonney2k | gsomix, exactly now that you are back CSet,CMap,Array/load/save etc are waiting | 20:25 |
* sonney2k is getting hungry | 20:25 | |
blackburn | sonney2k: what about borsch? | 20:25 |
blackburn | you should like borsch too :D | 20:26 |
gsomix | or solyanka | 20:26 |
@sonney2k | I love soljanka | 20:26 |
@sonney2k | though I don't recall the taste of borsch | 20:26 |
n4nd0 | blackburn: gsomix: do you know guys that this is really famous in Spain? http://en.wikipedia.org/wiki/Russian_salad | 20:26 |
blackburn | n4nd0: ooh that's nice salad sure | 20:27 |
n4nd0 | we call it ensaladilla rusa - Russian salad | 20:27 |
blackburn | n4nd0: http://en.wikipedia.org/wiki/Dressed_herring this one is very popular here too | 20:27 |
gsomix | n4nd0, we cook it for the New Year. whole buckets. :) | 20:28 |
blackburn | sonney2k: do you like selyodka pod shuboy? | 20:28 |
@sonney2k | I don't know what this is | 20:28 |
blackburn | sonney2k: http://en.wikipedia.org/wiki/Dressed_herring | 20:28 |
n4nd0 | I like vegetables, I think I would like Russian food | 20:28 |
@sonney2k | blackburn, how should I mark places where SGVector no longer works as expected? | 20:28 |
@sonney2k | TODOVEC ? | 20:28 |
blackburn | sonney2k: TODO? | 20:28 |
blackburn | heh TODO SGVECTOR or so | 20:29 |
@sonney2k | there are tons of these | 20:29 |
blackburn | oh that would be painful.. | 20:29 |
@sonney2k | it will require internal changes to machines, features, whatnot | 20:29 |
@sonney2k | I can type TODOVEC faster :) | 20:30 |
blackburn | hehe | 20:30 |
blackburn | ok | 20:30 |
blackburn | penetration is the word describing these processes | 20:31 |
@sonney2k | blackburn, btw I think we need some base class for GC stuff | 20:31 |
blackburn | GC? | 20:31 |
@sonney2k | garbage collection | 20:31 |
blackburn | sonney2k: yeah I understand but ? | 20:31 |
blackburn | GCwhat? | 20:31 |
@sonney2k | because sgvector, sgmatrix, etc all have this refcounting mechanism | 20:31 |
@sonney2k | no idea | 20:31 |
blackburn | what should it share? | 20:32 |
blackburn | ah I had one idea walking from job | 20:32 |
@sonney2k | SGRefObject? | 20:32 |
blackburn | we should implement clone() method based on parameters | 20:32 |
@sonney2k | ref/unref mechanism | 20:32 |
@sonney2k | with one virtual method to actually destroy things | 20:32 |
blackburn | sonney2k: what is the difference between SGRef and SG objects? | 20:32 |
@sonney2k | SGObject will derive from them | 20:33 |
@sonney2k | SGObjects can be serialized in addition to having refcounting | 20:33 |
blackburn | I lack Serialiazable interface | 20:33 |
@sonney2k | dump(blackburn) | 20:33 |
@sonney2k | not_implemented_error | 20:33 |
blackburn | heh | 20:34 |
blackburn | Serializable* | 20:34 |
@sonney2k | ? | 20:34 |
@sonney2k | you want SGObject and derive SGSerializable from it? | 20:34 |
blackburn | sonney2k: no I really like java interfaces | 20:34 |
@sonney2k | instead of SGRef -> SGObject , ... | 20:34 |
blackburn | noonono | 20:35 |
@sonney2k | I have no idea why c++0x doesn't have interfaces | 20:35 |
@sonney2k | that really is the only cool thing about java I've learned so far | 20:35 |
blackburn | sonney2k: the only?? | 20:35 |
blackburn | sonney2k: annotations? | 20:36 |
@sonney2k | blackburn, actually I like the idea of having SGObject -> SGSerializable | 20:36 |
blackburn | for each | 20:36 |
@sonney2k | never used this no idea what this is for | 20:36 |
@sonney2k | lame iterators | 20:36 |
@sonney2k | but c++0x has them | 20:36 |
blackburn | annotations? | 20:36 |
blackburn | sonney2k: do you know what dependency injection is? | 20:37 |
@sonney2k | seems like we are trapped in some loop | 20:37 |
@sonney2k | no | 20:37 |
@sonney2k | blackburn, focus man focus | 20:37 |
blackburn | haha | 20:37 |
blackburn | soryy | 20:37 |
@sonney2k | <sonney2k> blackburn, actually I like the idea of having SGObject -> SGSerializable | 20:37 |
@sonney2k | ^ | 20:37 |
@sonney2k | blackburn, | 20:37 |
blackburn | sonney2k: yes seems to be a good idea | 20:37 |
blackburn | sonney2k: actually we can imitate interfaces.. | 20:38 |
@sonney2k | blackburn, that will require a rename from SGObject -> SGSerializable first | 20:38 |
@sonney2k | blackburn, how so? | 20:38 |
blackburn | multiple inheritance | 20:38 |
@sonney2k | MI can do much more bad stuff yes | 20:38 |
blackburn | sonney2k: we won't have diamond there | 20:38 |
@sonney2k | but other stuff | 20:39 |
blackburn | as SGSerializable should only indicate class is serializable | 20:39 |
@sonney2k | no MI | 20:39 |
n4nd0 | I read once an article about simulating interfaces in C++ | 20:39 |
n4nd0 | I like interfaces too, maybe we could take a look to that article | 20:40 |
@sonney2k | as long as it doesn't involve MI - yeah | 20:41 |
blackburn | btw | 20:41 |
n4nd0 | well it uses some macros, but it MI indeed | 20:41 |
blackburn | don't you think we should have NAME("name") macros? | 20:41 |
n4nd0 | #define implements public | 20:41 |
blackburn | for get_name method | 20:41 |
n4nd0 | :P | 20:41 |
blackburn | hahahah | 20:41 |
blackburn | #define true false | 20:42 |
n4nd0 | don't get me wrong guys, I was not joking | 20:42 |
n4nd0 | the method had some other things | 20:42 |
blackburn | n4nd0: yes I understand | 20:42 |
n4nd0 | but at the end it relies on MI | 20:42 |
n4nd0 | :) | 20:42 |
blackburn | I am not big fan of MI too | 20:42 |
n4nd0 | I realized that it was a bit as I was pulling your legs :D | 20:43 |
blackburn | :D | 20:43 |
@sonney2k | blackburn, no idea how that can work but yes I wanted that too | 20:49 |
blackburn | sonney2k: wanted what? | 20:49 |
@sonney2k | NAME macros for the classes | 20:51 |
blackburn | should be easy? | 20:51 |
@sonney2k | how? | 20:51 |
blackburn | #define NAME(CLASSNAME) virtual const char* get_name() const { return CLASSNAME; } | 20:52 |
blackburn | won't it work? | 20:52 |
@sonney2k | blackburn, ahh yes | 20:52 |
@sonney2k | I wanted more | 20:52 |
blackburn | I tried to add GETTER/SETTER but failed | 20:52 |
blackburn | I guess it is not possible | 20:52 |
@sonney2k | that one writes this instead of class CFooBar { char* getrname() { return "FooBar"; } }; | 20:53 |
@sonney2k | blackburn, should sgvector.vector still be public? | 20:54 |
@sonney2k | and length? | 20:54 |
blackburn | yes | 20:54 |
blackburn | hmm | 20:54 |
blackburn | no | 20:54 |
blackburn | or yes | 20:54 |
blackburn | :D | 20:54 |
@sonney2k | hahahahah | 20:54 |
blackburn | sonney2k: ok I think no | 20:54 |
@sonney2k | yes is the easy answer so lets go for it for now | 20:54 |
blackburn | but would need to do some transition | 20:54 |
@sonney2k | even more than we have currently | 20:54 |
blackburn | I'd rather hide it | 20:55 |
blackburn | in future | 20:55 |
blackburn | :D | 20:55 |
-!- n4nd0 [02893bbe@gateway/web/freenode/ip.2.137.59.190] has quit [Ping timeout: 245 seconds] | 20:57 | |
blackburn | okay I have to study I think | 20:57 |
blackburn | I haven't read a page yet and exam is in 11 hours :D | 20:58 |
blackburn | however brb | 20:58 |
-!- blackburn [~qdrgsm@188.168.4.67] has quit [Quit: Leaving.] | 20:58 | |
@sonney2k | *lol* | 21:01 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 21:02 | |
@sonney2k | blackburn, 4 minutes learning? | 21:16 |
blackburn | sonney2k: hah no I'm reading that shit | 21:18 |
gsomix | 4 minutes learning? | I'm not surprised, blackburn studies all very fast. :) | 21:21 |
@sonney2k | gsomix, I am also no longer surprised | 21:28 |
blackburn | yeah I just imitate I know something :D | 21:28 |
@sonney2k | thats how you do it! | 21:28 |
@sonney2k | now I finally know | 21:28 |
blackburn | sonney2k: exactly | 21:28 |
* sonney2k dies in SGVector hell | 21:36 | |
blackburn | sonney2k: trying ecoc on road signs.. | 21:40 |
blackburn | yeah I can't focus on politics stuff | 21:40 |
-!- vikram360 [~vikram360@117.192.183.45] has quit [Ping timeout: 246 seconds] | 21:53 | |
-!- vikram360 [~vikram360@117.192.166.235] has joined #shogun | 21:54 | |
blackburn | hmm no way - sparse one is not working yet in a proper way :( | 21:57 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 22:08 | |
* sonney2k suffer**2 | 22:16 | |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has joined #shogun | 22:16 | |
gsomix | sonney2k, why we need inline T get_element(int32_t index) const in CSet? | 22:17 |
@sonney2k | gsomix, IIRC element() is for r/w access get_element not | 22:19 |
@sonney2k | if it is not needed -> delete it :) | 22:19 |
@sonney2k | (or not used yet) | 22:19 |
gsomix | sonney2k, ok, I figured out. | 22:22 |
* gsomix is fighting for bytes. | 22:23 | |
@sonney2k | I have like zillions of errors left | 22:27 |
blackburn | sonney2k: when I was first year undergraduate we were discussing with my friend how to compute vandermond matrix determinant and we decided to compute it step by step - and once we get N-5 x N-5 matrix we realized.. :D | 22:32 |
blackburn | so it should be something like N-353 errors left :D | 22:33 |
@sonney2k | blackburn, what is cool about the new sgvectro stuff is that you can write SGVector a = SGVector(...) | 22:39 |
@sonney2k | and get automagic refcount increase of a | 22:39 |
@sonney2k | and when the function exits a will be destroyed | 22:39 |
blackburn | sonney2k: nice, copy constructor? | 22:39 |
@sonney2k | so no need to manually call ref() / unref() | 22:39 |
@sonney2k | yes | 22:39 |
blackburn | awesome | 22:39 |
@sonney2k | there are some rare cases where you want to immediately kill the vector (call unref) | 22:40 |
@sonney2k | but usually not | 22:40 |
@sonney2k | it compiles! | 22:40 |
blackburn | yay | 22:40 |
@sonney2k | let me commit and break everything :( | 22:40 |
blackburn | haha | 22:40 |
wiking | nooooooooo | 22:42 |
wiking | anybody has any idea why latex creates an a4 dvi when i have \documentclass[10pt,twocolumn,letterpaper]{article}? :) | 22:42 |
blackburn | wiking: huh | 22:43 |
blackburn | why do you need dvi btw? | 22:43 |
CIA-113 | shogun: Evgeniy Andreev master * r61b244b / src/shogun/lib/DataType.h : added reference counting for SGVector (further SGIVector) - http://git.io/CNsNfw | 22:43 |
CIA-113 | shogun: Evgeniy Andreev master * r3f444b9 / src/shogun/lib/DataType.h : minor fixes - http://git.io/olNAiQ | 22:43 |
CIA-113 | shogun: Evgeniy Andreev master * r4b3c4f6 / src/shogun/lib/DataType.h : added no_ref_count arg, temporary changes - http://git.io/9maQJQ | 22:43 |
CIA-113 | shogun: Evgeniy Andreev master * r4fbd7c5 / src/shogun/lib/DataType.h : minor fixes 2 - http://git.io/9N9q4g | 22:43 |
wiking | well don't get me started with ieee :) | 22:43 |
@sonney2k | its done | 22:44 |
@sonney2k | the world will end | 22:44 |
@sonney2k | now | 22:44 |
@sonney2k | I better get some sleep | 22:44 |
blackburn | sonney2k: why gsomix name there? | 22:44 |
blackburn | :D | 22:44 |
@sonney2k | wiking, shouldn't it be letter? | 22:44 |
wiking | sonney2k: yes it should be, but it creates a4 :) | 22:44 |
wiking | and i have nooo idea why | 22:44 |
blackburn | hah right letter | 22:44 |
@sonney2k | blackburn, hmmhh I merged his stuff but added my stuff | 22:45 |
@sonney2k | not letterpaper | 22:45 |
@sonney2k | but letter | 22:45 |
wiking | and i would need letter :) | 22:45 |
blackburn | wiking: [10pt,twocolumn,letter]? | 22:45 |
@sonney2k | noooo | 22:46 |
@sonney2k | fuuuuu!K!J#@(*$#Y | 22:47 |
@sonney2k | all changes gone?! | 22:47 |
wiking | blackburn: hop does not work | 22:47 |
@sonney2k | blackburn, hmmhh seems like git rebase killed it all | 22:47 |
blackburn | sonney2k: did you use amend? | 22:47 |
@sonney2k | yes | 22:48 |
@sonney2k | and then git pull --rebase | 22:48 |
@sonney2k | git commit --amend -a | 22:48 |
@sonney2k | followed by git pull --rebase | 22:48 |
@sonney2k | after that all my changes were gone | 22:48 |
blackburn | huh | 22:48 |
@sonney2k | which is just 4 hours of work... | 22:48 |
blackburn | why did you pulled? | 22:48 |
@sonney2k | because there was some merge inbetween | 22:49 |
blackburn | is local repo broken too? | 22:49 |
blackburn | rebase shouldn't do that afaik | 22:50 |
@sonney2k | blackburn, you think it is somewhere in git history? | 22:50 |
@sonney2k | maybe it is due to this merge commit? | 22:50 |
blackburn | sonney2k: so is it lost in local repo too? | 22:50 |
@sonney2k | I don't see the commit... | 22:51 |
blackburn | sonney2k: hmm so where is your amended changes? | 22:53 |
blackburn | I do not believe it can disappear because of rebase | 22:53 |
blackburn | wiking: any idea? | 22:54 |
wiking | hehehe | 22:54 |
wiking | well rebase just rebases | 22:54 |
wiking | it's all gone mate | 22:54 |
@sonney2k | from the git faq | 22:54 |
@sonney2k | So, we would normally just git pull --rebase to get ready to push, but if we do that, the merge commit gets deleted! | 22:54 |
wiking | ideally | 22:54 |
@sonney2k | great | 22:55 |
blackburn | sonney2k: damn so did you amend merge commit? | 22:55 |
@sonney2k | for merge commits one should do git pull --rebase -p | 22:55 |
@sonney2k | yes | 22:55 |
@sonney2k | --preserve-merges | 22:55 |
@sonney2k | Instead of ignoring merges, try to recreate them. | 22:55 |
blackburn | one of the reasons I do not like amend | 22:56 |
@sonney2k | blackburn, not related to amend | 22:56 |
@sonney2k | related to merge commits | 22:56 |
@sonney2k | if you do a rebase on a merge commit your merge commit is gone | 22:56 |
shogun-buildbot | build #727 of r_static is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/727 blamelist: gsomix@gmail.com | 22:57 |
blackburn | sonney2k: but if you did basic commit it would be here still | 22:57 |
@sonney2k | great wasted this evening ... anyway need to sleep now | 22:57 |
@sonney2k | yes | 22:57 |
blackburn | bad :( | 22:58 |
blackburn | see you.. | 22:58 |
@sonney2k | git reset --hard origin/master would have recovered fromt his if I didn't push before *sigh* | 22:58 |
@sonney2k | btw one could also say rebase is bad | 23:00 |
@sonney2k | a normal pull would have been no problem | 23:00 |
@sonney2k | ohh well we will all learn the hard way | 23:00 |
blackburn | sonney2k: pull is bad | 23:00 |
shogun-buildbot | build #749 of cmdline_static is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cmdline_static/builds/749 blamelist: gsomix@gmail.com | 23:01 |
blackburn | fetch/{merge,rebase} is generally better controlled | 23:01 |
@sonney2k | blackburn, if you don't expect a problem then that would not have helped either | 23:01 |
blackburn | sure | 23:02 |
blackburn | sonney2k: painful wisdom - "never amend merge and rebase" | 23:03 |
shogun-buildbot | build #715 of python_static is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/python_static/builds/715 blamelist: gsomix@gmail.com | 23:05 |
@sonney2k | yeah that is the downside of using git - you can loose everything ... with svn this is not possible | 23:07 |
blackburn | sonney2k: we would break the neck doing that kind of development in svn | 23:10 |
-!- blackburn [5bde8018@gateway/web/freenode/ip.91.222.128.24] has quit [Ping timeout: 245 seconds] | 23:18 | |
@sonney2k | gsomix, do you remember what you changed to support SGVector& ? | 23:40 |
@sonney2k | I would like to get rid of this & stuff again... | 23:40 |
wiking | maaaan | 23:41 |
wiking | it took me this much! | 23:41 |
wiking | but made it | 23:41 |
wiking | fuck | 23:41 |
wiking | u killed it? | 23:42 |
gsomix | sonney2k, only arguments of methods and typemaps (added for SGVector&). | 23:42 |
@sonney2k | wiking, ?! | 23:44 |
@sonney2k | who did you kill? | 23:44 |
@sonney2k | gsomix, ok | 23:44 |
@sonney2k | gsomix, are you making progress? | 23:44 |
gsomix | sonney2k, in CSet and map to titled features? | 23:45 |
@sonney2k | well CSet first :) | 23:45 |
gsomix | yep, of course. I figured out how to make hashset with indexes. | 23:46 |
gsomix | unfortunately, it needed little more memory, than original CSet | 23:47 |
gsomix | but it should works faster. | 23:48 |
gsomix | *it need | 23:48 |
wiking | sonney2k: have u killed the git repo :) | 23:49 |
@sonney2k | wiking, ? | 23:50 |
wiking | sonney2k: the rebase... have u lost some commits? | 23:51 |
@sonney2k | all | 23:51 |
wiking | all?? | 23:51 |
wiking | shit | 23:51 |
wiking | :) | 23:51 |
-!- vikram360 [~vikram360@117.192.166.235] has quit [Ping timeout: 260 seconds] | 23:53 | |
-!- vikram360 [~vikram360@117.192.167.99] has joined #shogun | 23:53 | |
gsomix | sonney2k, btw, can you explain me later how to make templates with two args in shogun? | 23:54 |
wiking | template<class this, class that> | 23:54 |
wiking | ? | 23:54 |
gsomix | wiking, I have some problems with class_list.cpp | 23:55 |
wiking | or typename instead of class... depends what u want to template | 23:55 |
wiking | what the error | 23:55 |
wiking | *what's | 23:55 |
wiking | gsomix copy-paste me a gcc error! | 23:58 |
gsomix | wiking, related to declaration templated classes with standart types. It seems, shogun thinks, that all templated classes have only one arg. | 23:58 |
gsomix | wiking, later. ok? | 23:58 |
wiking | hehehe ok ok | 23:59 |
gsomix | templated SG classes, I mean | 23:59 |
wiking | i might not be here later though | 23:59 |
--- Log closed Thu May 03 00:00:37 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!