| --- Log opened Mon Apr 18 00:00:36 2011 | ||
| serialhex | :-/ everyones probably asleep... well maybe not alesis-novik, i'm not sure if he does sleep :D | 01:16 |
|---|---|---|
| serialhex | well, maybe he is...? | 01:42 |
| -!- josip [~josip@95.180.215.37] has joined #shogun | 01:46 | |
| -!- josip [~josip@95.180.215.37] has quit [Changing host] | 01:46 | |
| -!- josip [~josip@unaffiliated/josip] has joined #shogun | 01:46 | |
| alesis-novik | serialhex, he's not | 02:27 |
| serialhex | nice to see you | 02:27 |
| serialhex | alesis-novik, so whats going on? | 02:34 |
| alesis-novik | serialhex, working on EM for GMM for tomorrow | 02:41 |
| serialhex | cool | 02:41 |
| serialhex | tomorrow's the day we fiund out who is voted off the island! | 02:42 |
| alesis-novik | who was the weakest link | 02:43 |
| serialhex | yes indeed! | 02:43 |
| serialhex | though it seems ~50 people have already dropped out of the running, we've still got 7 or so people & only 5 slots | 02:45 |
| alesis-novik | maybe a few more, like ~10 | 02:47 |
| josip | serialhex: what voring | 02:55 |
| josip | voting* | 02:55 |
| josip | i thought the final decisions come a bit later this week | 02:55 |
| serialhex | afaik soeren & the rest of the mentors are getting together tomorrow & deciding whos doing what | 02:56 |
| serialhex | (soeren == sonney2k btw) | 02:56 |
| josip | oh, okay | 02:56 |
| josip | but we still won't find out until friday | 02:56 |
| serialhex | now my information could be wrong, but i read it somewhere in the chat, so we'll see | 02:57 |
| josip | or rather next monday | 02:57 |
| serialhex | maybe, maybe not? google may not 'officially' inform us but soeren may unofficially tell us whats up | 02:57 |
| josip | I don't think they're instructed to do that | 02:58 |
| serialhex | i'm not sure... but what i _am_ sure about is firefly is on right now... ttyl!! | 03:02 |
| josip | hehe, enjoy it | 03:02 |
| -!- Ziyuan [~Ziyuan@116.23.212.45] has quit [] | 03:07 | |
| alesis-novik | How many times have you seen it by now serialhex? :D | 03:20 |
| serialhex | alesis-novik: the whole series? none, but i have watched various episodes a couple of times... they're airing them aain on the science channel, so i get to watch the series from the beginning now!!! :D | 03:20 |
| alesis-novik | and yes, the mentors have to assign their slots tomorrow, but then later this week there will be duplication elimination so the *official* results are only next monday | 03:21 |
| serialhex | (btw, i LOVE dvr's!!!) | 03:21 |
| serialhex | ahh, i see | 03:21 |
| alesis-novik | so serialhex, very important question... BSG vs Firefly? :D | 03:22 |
| serialhex | oooh, a toughie! | 03:22 |
| serialhex | i'll have to go with BSG, only because it's finished... though they are both up there on my 'shows to watch' list! | 03:23 |
| alesis-novik | good answer. I'll be back in ~30 mins | 03:24 |
| * serialhex goes to rot his brain more :D | 03:27 | |
| josip | alesis-novik: but students won't have an idea of what's going on until Monday, correct? | 03:31 |
| josip | sonney2k: I am looking at the evaluation code, I was travelling accross Europe for the weekend :D | 03:31 |
| josip | s/accross/across* | 03:31 |
| -!- josip [~josip@unaffiliated/josip] has quit [Ping timeout: 240 seconds] | 03:42 | |
| -!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 03:46 | |
| -!- josip [~josip@unaffiliated/josip] has joined #shogun | 03:53 | |
| josip | did I miss anything? I plugged my computer off involuntarily | 03:54 |
| serialhex | josip: nope, you didnt miss anything | 03:55 |
| josip | (sicne my last message) | 03:55 |
| serialhex | nope | 03:56 |
| -!- siddharth [~siddharth@117.211.88.150] has quit [Read error: Connection reset by peer] | 04:08 | |
| -!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 04:12 | |
| siddharth | hi all | 04:12 |
| serialhex | mornin siddharth! | 04:38 |
| siddharth | morning serialhex | 04:38 |
| serialhex | it's actually 10:40pm here :P so more like evening :P | 04:39 |
| siddharth | good night serialhex :P | 04:39 |
| @bettyboo | ;D | 04:39 |
| siddharth | hi bettyboo | 04:39 |
| serialhex | lol, i'm watching tv & stuff, so no bed for a little while at least | 04:39 |
| @bettyboo | siddharth: oh.. sorry | 04:40 |
| siddharth | which tv show? | 04:40 |
| serialhex | watching punisher (movie) right now & then i'm going to watch firefly - with no commercials!!! (thank god for DVR's!!!!!) | 04:41 |
| siddharth | firefly has only 2 seasons right? | 04:42 |
| serialhex | i think so... they never finished the series, but it's still an awesome show! | 04:43 |
| siddharth | ok will watch it someday | 04:45 |
| siddharth | are there anymore sci-fi tv series? | 04:45 |
| serialhex | ehhh... nothing new that i'm aware of unfortunately. maybe Dr. Who (which has a new season coming out soon, over here anyway) but i've never seen it before so idk | 04:49 |
| serialhex | ...anyway, back to pizza && television! | 04:49 |
| siddharth | enjoy :) | 04:51 |
| * siddharth back to work | 04:51 | |
| alesis-novik | Firefly only has 1 season | 04:55 |
| alesis-novik | and then there's Serenity | 04:55 |
| -!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 05:04 | |
| Tanmoy | mornin | 05:04 |
| Tanmoy | every1 | 05:04 |
| alesis-novik | Morning Tanmoy | 05:05 |
| alesis-novik | And the answer is yes, I do not sleep. Ever. | 05:05 |
| Tanmoy | k nice | 05:08 |
| Tanmoy | so alesis does the GHMM seem to be working or not | 05:26 |
| alesis-novik | GHMM? no idea, I wasn't the one working on that Tanmoy | 05:35 |
| -!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 05:37 | |
| Tanmoy | yeah i knw just tht u might have some idea :) | 05:40 |
| alesis-novik | No, sorry | 05:46 |
| alesis-novik | Right, I'm done for today | 07:21 |
| alesis-novik | See you today folks | 07:21 |
| -!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Read error: Connection reset by peer] | 07:25 | |
| -!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 07:26 | |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 07:40 | |
| -!- lionelc [4c681efd@gateway/web/freenode/ip.76.104.30.253] has joined #shogun | 08:07 | |
| -!- blackburn [~qdrgsm@109.226.117.183] has joined #shogun | 09:00 | |
| blackburn | morgen | 09:03 |
| blackburn | alesis-novik: рвёшь и мечешь ;) | 09:03 |
| @sonney2k | blackburn, good morning | 09:09 |
| @sonney2k | I am tempted to merge your evaluation changes just now already | 09:09 |
| @sonney2k | they look like a big improvement | 09:09 |
| @sonney2k | over what we have now | 09:09 |
| blackburn | hmm.. you could but I have a problem with BAL :) | 09:10 |
| blackburn | last night I was confused and wrote some non-sense | 09:10 |
| blackburn | sonney2k: can you help me with it? | 09:10 |
| @sonney2k | blackburn, I actually just looked at it and it looks wrong indeed | 09:11 |
| blackburn | pidx=(lab==+1) | 09:11 |
| blackburn | perr=numpy.mean(out[pidx]>0) | 09:11 |
| blackburn | it is from mldara utils | 09:11 |
| blackburn | mldata | 09:11 |
| @sonney2k | just compute the error on the postive examples and the negtaively labeled examples | 09:11 |
| @sonney2k | then average | 09:11 |
| blackburn | but is it error? I mean we check error of positive on idxs where positive should be? | 09:12 |
| Tanmoy | @sonney2k does the GHMM need a new implementation | 09:13 |
| Tanmoy | i am tempted and have been doing some stuff | 09:13 |
| blackburn | sonney2k: if ground_truth[idx]>1 but predicted[idx]<1 then pos_err++, right? | 09:13 |
| @sonney2k | blackburn, ohh true looks wrong | 09:15 |
| blackburn | sonney2k: so seems that in mldata it isn't true, right? | 09:16 |
| @sonney2k | blackburn, yes | 09:17 |
| blackburn | hehe. | 09:17 |
| @sonney2k | blackburn, look at http://sonnenburgs.de/soeren/publications/Son08.pdf section 5.1 (page 72 - table 5.2) | 09:20 |
| @sonney2k | I have a summary table there | 09:20 |
| blackburn | sonney2k: thank you | 09:20 |
| @sonney2k | blackburn, my suggestion would be to (like in the current perf measures) have helper functions to compute fp/tp/fn/tn and then from it most of the scaler scors | 09:21 |
| @sonney2k | scores | 09:21 |
| @sonney2k | except for auROC / auPRC | 09:21 |
| blackburn | sonney2k: seems to be a very good suggestion ;) | 09:22 |
| @sonney2k | blackburn, use float64_t for counting - just to be safe :) | 09:23 |
| blackburn | okay | 09:23 |
| Tanmoy | @sonney2k does the GHMM needs a new imp | 09:24 |
| @sonney2k | Tanmoy, yes there is none so far (GHMM == generalized hmm). but this is a long term project right? | 09:25 |
| Tanmoy | well i thought it wld be use for HM-SVM | 09:26 |
| Tanmoy | i meant Semi HM SVM | 09:26 |
| Tanmoy | on a long term wat really wld be nice wld be be to integrate it | 09:27 |
| Tanmoy | http://cs.ru.nl/~jorism/libDAI/ | 09:27 |
| Tanmoy | so u have a package on even Graphical Models | 09:28 |
| Tanmoy | oh to add for structured learning too | 09:29 |
| @sonney2k | Tanmoy, yes, but merging with such big projects is difficult. | 09:30 |
| Tanmoy | well not now did for an overall effort probably in parts | 09:31 |
| Tanmoy | libdai has some LP progr as parts of inference in structured learning which could be useful | 09:32 |
| Tanmoy | oh so i tht GHMM is only for Semi Markov models | 09:32 |
| @sonney2k | yes | 09:40 |
| @mlsec | decision Day! | 09:42 |
| blackburn | hehe | 09:44 |
| blackburn | sonney2k: I have an idea about ROC and other graphs | 09:45 |
| blackburn | sonney2k: evaluate should return for example auROC, and evaluate_graph will return ROC | 09:45 |
| @sonney2k | blackburn, lets see if I have the same :) | 09:45 |
| @sonney2k | yes | 09:45 |
| @sonney2k | great! | 09:45 |
| @sonney2k | twins in mind or so ;-) | 09:45 |
| blackburn | :D | 09:45 |
| blackburn | sonney2k: there is a lot of rates in that table.. | 09:46 |
| blackburn | do we need all of them? | 09:47 |
| blackburn | I will have no problem with method scoring tp,tn,..., but the number of classes will be >15 :) | 09:47 |
| @sonney2k | blackburn, since computing auROC is as expensive as computing ROC (more or less) I would return the ROC score on the evaluation graph | 09:48 |
| blackburn | the only problem I have - evaluate of Evaluation returns float64_t but not float64_t* | 09:49 |
| blackburn | that's why I can not return graph from evaluate() | 09:49 |
| @sonney2k | blackburn, doesn't matter - I mean each class is just one function. the only alternative is to put things in a one class - that is rather not so obvious way (we do this though too) | 09:49 |
| blackburn | sonney2k: but what we shall do with ROC? | 09:50 |
| blackburn | Evaluation will not fit for classes computing graphs.. | 09:50 |
| @sonney2k | blackburn, why not do it like this then: float64_t evalutation_graph(float64_t* &graph, int32_t &graph_len) | 09:50 |
| blackburn | yeap, I mean it too | 09:50 |
| @sonney2k | blackburn, why not? | 09:51 |
| blackburn | having evaluation_graph it will work | 09:51 |
| @sonney2k | each graph has an associated scalar measure | 09:51 |
| blackburn | but what will do evaluate()? | 09:51 |
| @sonney2k | compute the area under that graph only | 09:51 |
| @sonney2k | so return what evaluation_graph returns (but not the graph) | 09:52 |
| blackburn | ok | 09:52 |
| blackburn | and if evaluate will be called at first it could make a graph | 09:53 |
| blackburn | and 'remember' it somewhere for future calls of evaluation_graph | 09:53 |
| blackburn | when will you have 'phone conf' about us? :D | 09:57 |
| @sonney2k | in 5hrs | 09:57 |
| @sonney2k | blackburn, it could but I don't know if it is really necessary. | 09:58 |
| blackburn | anyway, now I'm doing BAL and etc | 09:59 |
| @sonney2k | blackburn, I mean one rarely wants all of the eval scores | 09:59 |
| blackburn | ah | 09:59 |
| @sonney2k | if so - it would make a lot of sense to have the design in one class - and compute TPR/etc there and then just return things like TP+TN/(TP+TN+FN+FP) | 10:01 |
| blackburn | sonney2k: but don't know how to dispatch it: what will evaluate() evaluate and how can we choose it | 10:02 |
| -!- meerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 10:04 | |
| @sonney2k | would this work with static variables/ functions (not sure here)? I mean when one computes e.g. accuracy TPR/... are created and then one could get the BAL for free? | 10:04 |
| @sonney2k | but honestly I never needed this... | 10:04 |
| @sonney2k | so I am not sure if it is even worth the thought | 10:04 |
| @sonney2k | maybe putting all the contingency table based measures into one class would work and then in the constructor one selects which one should be used upon evaluate() call | 10:05 |
| blackburn | yeap, that idea seems to be good | 10:06 |
| -!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 276 seconds] | 10:06 | |
| blackburn | but how we will choose it, with it's name? | 10:06 |
| @sonney2k | blackburn, some enum again | 10:07 |
| blackburn | sonney2k: but how we will use it in modular? | 10:07 |
| blackburn | can we make enums visible in e.g. python? | 10:07 |
| @sonney2k | CContingencyTableEvaluation(ACCURACY) | 10:07 |
| @sonney2k | blackburn, yes | 10:07 |
| @sonney2k | they are just integers there | 10:07 |
| -!- meerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 250 seconds] | 10:08 | |
| blackburn | okay | 10:08 |
| blackburn | but don't know what to 'redesign' | 10:08 |
| @sonney2k | the only problem I see is that it hides the name Accuracy ... so beginners may have problems finding it | 10:08 |
| blackburn | MulticlassAccuracy, MSE will stay | 10:10 |
| blackburn | but Accuracy and BAL will be replaced by some another 'base' class | 10:10 |
| @sonney2k | blackburn, I think the way you do it now is easier to digest having a contingencytableevalution class later on that just computes all the scores for performance can be done later (if ever needed - which I don't think will be the case) | 10:11 |
| @sonney2k | blackburn, that is good | 10:11 |
| @sonney2k | even better idea | 10:11 |
| @sonney2k | then it works | 10:11 |
| blackburn | eh.. so don't change it now? | 10:12 |
| blackburn | sonney2k: I fixed BalancedError with get_scores() | 10:16 |
| blackburn | for TP, etc | 10:16 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 10:16 | |
| -!- siddharth [~siddharth@117.211.88.150] has quit [Ping timeout: 240 seconds] | 10:17 | |
| -!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 10:21 | |
| blackburn | sonney2k: another idea: create a base class with that get_scores() and fields for TP, FN, ... and describe all the evaluators inline | 10:25 |
| @sonney2k | blackburn, got disconnected, did you still see the: blackburn, I mean I like the idea with the CContingencyTable class derived from CEvaluation / then it is fast and transparent! | 10:43 |
| blackburn | sonney2k: I propose slightly different, we could make TwoClassEvaluation base class | 10:44 |
| blackburn | and then derives Accuracy, etc | 10:44 |
| blackburn | how about this way? | 10:45 |
| @sonney2k | BinaryClassEvaluation - but then you still need to derive the contingencytableevaluaton class from that | 10:46 |
| @sonney2k | right? | 10:49 |
| blackburn | a moment | 10:49 |
| blackburn | sonney2k: I will show you how it done now | 10:49 |
| @sonney2k | ok | 10:50 |
| blackburn | https://github.com/shogun-toolbox/shogun/pull/51/files | 10:51 |
| blackburn | sonney2k: take a look on BinaryClassEvaluation | 10:51 |
| blackburn | and Accuracy | 10:51 |
| @sonney2k | blackburn, one can compute accuracy/ roc etc also for multi-class | 10:54 |
| blackburn | MulticlassAccuracy stays | 10:55 |
| @sonney2k | so I think you need another class layer BinaryClassContingencyTable in there | 10:55 |
| @sonney2k | err but auROC is not derived from contingency table | 10:55 |
| blackburn | but why we need BinaryClassContingencyTable? | 10:55 |
| @sonney2k | blackburn, multiclass looks totally different... | 10:57 |
| blackburn | sonney2k: MulticlassAccuracy? | 10:59 |
| blackburn | or what you mean? | 10:59 |
| @sonney2k | blackburn, I think we have a misunderstanding here | 10:59 |
| blackburn | exactly :) | 10:59 |
| @sonney2k | base class is CEvaluation | 10:59 |
| blackburn | yeap | 10:59 |
| @sonney2k | then you derive MultiClassEvaluation / Binary etc | 11:00 |
| blackburn | not sure, now just BinaryClassEvaluation and MSE with MulticlassAccuracy | 11:00 |
| blackburn | btw, what is the ROC for multiclass?.. | 11:01 |
| @sonney2k | now for BinaryClassEvaluation siblings would be AreaUnderROCCurve and AreaUnderPRCurve and BinaryContingencyTableEvalution | 11:01 |
| @sonney2k | from BinaryContigencyTableEvaluation you could then derive Accuracy etc | 11:01 |
| blackburn | ah, I see | 11:02 |
| @sonney2k | blackburn, I would ignore more complex multiclass things for now | 11:02 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 11:02 | |
| blackburn | sonney2k: BinaryContingencyTableEvaluation seems to be a bit complex name.. may be BinaryClassRateEvaluation or etc? | 11:03 |
| @sonney2k | and since most of the code is binary class - I would for also drop the Binary prefix (just like we assume in Kernels that DotFeatures are used) | 11:04 |
| blackburn | well so let's decide naming | 11:05 |
| blackburn | base is CEvaluation, base class for evaluation 2-class is ..? | 11:05 |
| blackburn | *if not binary* | 11:06 |
| @sonney2k | it is BinaryClassEvaluation | 11:06 |
| blackburn | but you like to drop binary :) | 11:07 |
| @bettyboo | yep!?! | 11:07 |
| @sonney2k | I mean the subclasses like ContingencyTable / Accuracy ... that should not have the prefix | 11:07 |
| @sonney2k | yep | 11:07 |
| blackburn | ah | 11:07 |
| @sonney2k | there are ~20 of them right | 11:07 |
| blackburn | ah, okay | 11:08 |
| @sonney2k | so whenever the contingency table evaluate() function is called it computes all the other scores and one can easily access them via the get_accuracy etc functions | 11:11 |
| blackburn | eh.. that means we should not make Accuracy, etc | 11:12 |
| @sonney2k | so better rename the function get_scores to compute_scores | 11:12 |
| * blackburn is irritated by his catching cold | 11:12 | |
| @sonney2k | blackburn, no no | 11:12 |
| @sonney2k | keep them | 11:12 |
| blackburn | sonney2k: but if we will compute every score and use only one, we will lost all other | 11:13 |
| @sonney2k | e.g. CAccuracy's evaluate function calls CContingencyTable::evaluate() and returns get_accuracy() | 11:13 |
| @sonney2k | but anyone who needs more efficent access, can now also call get_BAl() etc | 11:13 |
| @sonney2k | blackburn, the contingency class stores them all | 11:14 |
| @sonney2k | I mean TP/FP/FN/TN that is | 11:14 |
| blackburn | so.. | 11:15 |
| blackburn | one can make: | 11:15 |
| blackburn | a = ContingencyTable() | 11:15 |
| blackburn | a.evaluate(labels1,labels2) | 11:15 |
| blackburn | a.get_BAL() | 11:15 |
| blackburn | a.get_accuracy() | 11:15 |
| blackburn | etc | 11:15 |
| blackburn | but same time can | 11:16 |
| blackburn | a = Accuracy() | 11:16 |
| blackburn | a.evaluate(labels1,labels2) | 11:16 |
| blackburn | sonney2k: is it? | 11:16 |
| -!- bettyboo [~bettyboo@bane.ml.tu-berlin.de] has quit [Ping timeout: 240 seconds] | 11:16 | |
| @sonney2k | blackburn, ahh you are right ContingencyTable's evaluate() is not defined | 11:17 |
| @sonney2k | so a = Accuracy() | 11:17 |
| @sonney2k | acc=a.evaluate(labels1,labels2) | 11:17 |
| @sonney2k | err=a.get_error() | 11:18 |
| @sonney2k | first call expensive - second cheap | 11:18 |
| blackburn | should accuracy compute error too? | 11:18 |
| -!- mlsec [~mlsec@bane.ml.tu-berlin.de] has quit [Ping timeout: 260 seconds] | 11:18 | |
| @sonney2k | blackburn, since you call CContingencyTable::compute_scores() you have all the tp/fp/... available | 11:19 |
| @sonney2k | no | 11:19 |
| blackburn | but you called it.. | 11:19 |
| blackburn | (01:18:08 PM) sonney2k: err=a.get_error() | 11:19 |
| @sonney2k | but you inherit the function from ContingencyTable | 11:19 |
| blackburn | ah! | 11:19 |
| @sonney2k | does that make sense? | 11:20 |
| blackburn | seems so | 11:20 |
| blackburn | but I don't sure we need derived classes like accuracy | 11:20 |
| blackburn | may be we could make only the most 'popular' ones | 11:21 |
| @sonney2k | blackburn, it is true, one could have a constructor CContingencyTable(E_CONT_PERF_MEASURE) and then let it's evaluate function return the right score | 11:24 |
| @sonney2k | it is just that I don't know if a ML beginner understands that (and is not looking for CAccuracry directly) | 11:24 |
| @sonney2k | hmmhh... just start with the contingency table class and the enum | 11:25 |
| blackburn | sonney2k: aha, working on it now | 11:26 |
| @sonney2k | then we could add these 2-line helper classes in one header file | 11:26 |
| @sonney2k | I mean class CAccuracy : public CContingencyTable { CAccuracy() : CContingencyTable(ACCURACY) { } } - thats all right | 11:27 |
| blackburn | and name | 11:28 |
| @sonney2k | and name (class name without C) - yes | 11:29 |
| @sonney2k | I think it is easy with a macro - and then one just names the 10-15 classes and done :) | 11:29 |
| blackburn | yeap | 11:30 |
| @sonney2k | (and even without macro it is...) | 11:30 |
| blackburn | sounds good | 11:30 |
| @sonney2k | I am happy with that plan too | 11:30 |
| @sonney2k | thanks for your work... | 11:30 |
| blackburn | no problem :) | 11:31 |
| blackburn | thanks for consulting me | 11:31 |
| blackburn | sonney2k: enum ERateType? | 11:34 |
| blackburn | or better EContingencyTableMeasureType? | 11:34 |
| @sonney2k | the latter looks better | 11:37 |
| @sonney2k | thanks | 11:37 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 11:46 | |
| CIA-110 | shogun: Ziyuan Lin master * r7f50fde / (4 files): Implement Bessel kernel - http://bit.ly/fxhlQ3 | 11:48 |
| CIA-110 | shogun: Soeren Sonnenburg master * rfec377d / (2 files): add bessel kernel thanks ziyuan for the patch - http://bit.ly/h2UGzb | 11:48 |
| blackburn | sonney2k: get_accuracy() or get_accuracy(CLabels* predicted, CLabels* ground_truth)? | 11:51 |
| blackburn | sonney2k: is it better to use set_type() for classifier and then use evaluation()? | 11:53 |
| blackburn | evaluate()* | 11:53 |
| blackburn | it's about doing get_accuracy(), ... private or protected | 11:56 |
| @sonney2k | blackburn, jsut get_accuracy() as poublic function | 11:57 |
| @sonney2k | assume that TP/... values are set | 11:57 |
| blackburn | fail if they aren't? | 11:58 |
| @sonney2k | blackburn, then you need one more boolean varialble that says they are inited | 11:58 |
| @sonney2k | gtg | 11:58 |
| @sonney2k | l8r | 11:58 |
| blackburn | see you | 11:58 |
| @sonney2k | re | 12:23 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 12:25 | |
| * blackburn wonders about deduplication: how they will deduplicate students if nobody of them will know anything for sure | 12:39 | |
| josip | you will be contacted if you're a top applicant for > 1 projects? | 12:40 |
| blackburn | seems so | 12:42 |
| blackburn | sonney2k: I removed Accuracy.h but have "No rule to make target `../shogun/Evaluation.h', needed by `Evaluation_wrap.cxx'. Stop." | 12:56 |
| blackburn | I did 'make clean' and removed it from Evaluation.i :( | 12:57 |
| blackburn | oh, not removed from evaluation | 12:58 |
| blackburn | earlier such error was with Accuracy | 12:58 |
| siddharth | sonney2k, pulled request for SGD-QN | 13:32 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 13:55 | |
| lionelc | Hi, sonney2k, I also sent a pull request about the parameter framework (for cross-validation). sorry it may be late... but really had an overwhelming week last week | 14:00 |
| @sonney2k | blackburn, you need to redo dependencies | 14:22 |
| @sonney2k | make depend IIRC | 14:23 |
| blackburn | sonney2k: all working now | 14:23 |
| @sonney2k | blackburn, :) | 14:23 |
| blackburn | anyway, thank you, will do it in case of that thing | 14:23 |
| * sonney2k loves it when the problems disappear | 14:23 | |
| blackburn | now have another problem :D | 14:23 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 14:24 | |
| blackburn | sonney2k: have problem with defining multiple classes in one header :) | 14:25 |
| @sonney2k | blackburn, good job :) | 14:26 |
| @sonney2k | forgot ; ? | 14:26 |
| -!- nicococo [~nicococo@lacedcoffee.ml.tu-berlin.de] has joined #shogun | 14:26 | |
| blackburn | sonney2k: i will commit it, may be you will see some mistake ;) | 14:26 |
| @sonney2k | ok | 14:27 |
| * blackburn wonders of 25 forks | 14:28 | |
| blackburn | sonney2k: https://github.com/shogun-toolbox/shogun/pull/51/files#diff-4 | 14:29 |
| -!- mlsec [~mlsec@bane.ml.tu-berlin.de] has joined #shogun | 14:29 | |
| -!- mode/#shogun [+o mlsec] by ChanServ | 14:29 | |
| -!- bettyboo [~bettyboo@bane.ml.tu-berlin.de] has joined #shogun | 14:29 | |
| -!- mode/#shogun [+o bettyboo] by ChanServ | 14:29 | |
| blackburn | oh, may it be a problem on my side? | 14:31 |
| -!- lionelc [4c681efd@gateway/web/freenode/ip.76.104.30.253] has quit [Quit: Page closed] | 14:35 | |
| @sonney2k | blackburn, I would put all the helper classes at the very bottom oh ContingencyTableEvaluation.h | 14:35 |
| -!- lionelc [4c681efd@gateway/web/freenode/ip.76.104.30.253] has joined #shogun | 14:35 | |
| blackburn | hm | 14:38 |
| blackburn | ok | 14:38 |
| blackburn | sonney2k: btw, 'make depend' isn't working | 14:38 |
| @sonney2k | blackburn, then make distclean and do the whole configure make again | 14:38 |
| blackburn | ok, thank you | 14:39 |
| blackburn | sonney2k: ahha, I forgot public in helpers... | 14:42 |
| blackburn | damn, I really have to get rid of cold I get :) | 14:43 |
| @sonney2k | blackburn, your patch really looks good - I had only minor comments as you can see | 14:47 |
| blackburn | sonney2k: I fixed it already, just have to test it now | 14:47 |
| @sonney2k | only ROC / PRC are missing now | 14:47 |
| blackburn | sonney2k: I will do it a bit later, ok? | 14:48 |
| @mlsec | betty, you left the channel without asking for permission? | 14:54 |
| @bettyboo | mlsec: you asking for me? | 14:54 |
| @mlsec | bettyboo: yes?! why did you leave? | 14:54 |
| @bettyboo | mlsec: also, there are no classes that provide general optimization methods, right? or did I miss something again? :) | 14:54 |
| @mlsec | bettyboo: yes, but aside from these classes. you cannot leave the channel without permission | 14:54 |
| @bettyboo | mlsec: betty, you left the channel without asking for permission? | 14:54 |
| @mlsec | bettyboo: very funny! | 14:54 |
| @bettyboo | mlsec: I did it in WaveKernel and CauchyKernel which are very similar to DistanceKernel | 14:54 |
| @mlsec | bettyboo: fortunately, you did not apply for gsoc. so we don't need to talk to you over and over again. | 14:57 |
| @bettyboo | mlsec: i'm on prettymuch the same schedule, i dont think i've gone to sleep before 1am in over a week | 14:57 |
| blackburn | bettyboo: will you marry serialhex? | 15:10 |
| @bettyboo | blackburn: pull is here | 15:10 |
| blackburn | bettyboo: where? | 15:10 |
| @bettyboo | blackburn: which is where? | 15:10 |
| blackburn | bettyboo: which watch? | 15:10 |
| @bettyboo | blackburn: http://www.youtube.com/watch?v=BLGWnaYKrdE&feature=player_detailpage#t=22s | 15:10 |
| blackburn | bettyboo: thank you | 15:11 |
| @bettyboo | blackburn: ah, you mean kernel trick with SVMs | 15:11 |
| @bettyboo | blackburn, what! | 15:11 |
| blackburn | bettyboo: let me know more about kernel trick | 15:12 |
| @bettyboo | blackburn: the idea said that there is a matlab implementation for EM and that the student should base the project on it. I know for a simple version of EM you need to be able to compute the PDF | 15:12 |
| @mlsec | well, i guess there is some confusion regarding EM and kernel trick. | 15:13 |
| @mlsec | bettyboo: you need to practice your skills a little more | 15:13 |
| @bettyboo | mlsec: yeah, you have a little more than 3 hours | 15:13 |
| @mlsec | bettyboo: YOU need to practice; not me. | 15:14 |
| @bettyboo | mlsec: *woe is all who do not get selected to be part of the summer of code* | 15:14 |
| blackburn | bettyboo: you are applied to GSoC! :D | 15:16 |
| @bettyboo | blackburn: 4 am, are you crazy? :D | 15:16 |
| blackburn | nice | 15:16 |
| @mlsec | bettyboo: but seriously, you gonna marry soon? | 15:19 |
| @bettyboo | mlsec: MLers who can code are few and far betwee | 15:19 |
| @mlsec | bettyboo: so you took the first you found. ;) | 15:19 |
| @bettyboo | mlsec: coulnd' you use jn() instead of the cyl_bessel_j ? | 15:19 |
| @mlsec | i see | 15:20 |
| -!- blackburn [~qdrgsm@109.226.117.183] has quit [Quit: Leaving.] | 15:25 | |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 15:32 | |
| @mlsec | "hier rockt das haus" | 15:53 |
| @sonney2k | bettyboo and mlsec rocken die hütte! | 15:55 |
| @bettyboo | sonney2k: so it boils down to selecting chunks to be optimized in the qp solver, selecting the chunks effecitvely using the alphas that have the highest gradient, doing speedups for linear svms and shrinking | 15:55 |
| siddharth | sonney2k, I have sent a pull request for the SGD-QN classifier | 16:00 |
| @mlsec | bettyboo: i will polish your backend now. relax and enjoy | 16:29 |
| @bettyboo | mlsec: lle, sne and isomap would also be interesting, as they all require a knn backend | 16:29 |
| -!- bettyboo [~bettyboo@bane.ml.tu-berlin.de] has quit [Remote host closed the connection] | 16:29 | |
| -!- bettyboo [~bettyboo@bane.ml.tu-berlin.de] has joined #shogun | 16:33 | |
| -!- mode/#shogun [+o bettyboo] by ChanServ | 16:33 | |
| @mlsec | welcome back, bettyboo | 16:33 |
| @bettyboo | mlsec: x.getwidth() | 16:33 |
| @mlsec | :> | 16:33 |
| * mlsec ...leftover cake in the kitchen!!! (^_^) | 16:37 | |
| -!- Ziyuan [~Ziyuan@116.21.255.137] has joined #shogun | 16:42 | |
| @mlsec | bettyboo: want some cake? | 16:42 |
| @bettyboo | mlsec: it takes the time in the future and optionally your latitue and the time you want now to be based of of, and calculates how far you travel (through space) | 16:42 |
| @mlsec | bettyboo: what? | 16:43 |
| @bettyboo | mlsec: that's what? | 16:43 |
| @mlsec | bettyboo: i don't get it | 16:43 |
| @bettyboo | mlsec: well to run simple things we don't even need typemaps | 16:43 |
| @mlsec | bettyboo: you are right. swig is cool. especially, with typemaps | 16:43 |
| @bettyboo | mlsec: @all going to bed, good night! | 16:44 |
| @mlsec | okay | 16:44 |
| @mlsec | nite | 16:44 |
| @mlsec | nicococo: alive? | 16:45 |
| nicococo | si senor | 16:45 |
| @mlsec | great. so, how's the cake? | 16:45 |
| nicococo | delicious.. my second piece already | 16:46 |
| CIA-110 | shogun: Soeren Sonnenburg master * r152a967 / (8 files): fix include paths to be relative to libshogun/ (+7 more commits...) - http://bit.ly/fLcefZ | 16:48 |
| -!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 16:48 | |
| -!- nicococo [~nicococo@lacedcoffee.ml.tu-berlin.de] has left #shogun [] | 16:57 | |
| -!- lionelc [4c681efd@gateway/web/freenode/ip.76.104.30.253] has quit [Quit: Page closed] | 17:01 | |
| siddharth | mlsec, so u have decided the 5 slots? | 17:01 |
| @mlsec | hmmm, maybe | 17:02 |
| siddharth | no last changes? | 17:02 |
| @mlsec | we have 4 +/- 1 | 17:02 |
| siddharth | i see | 17:02 |
| Ziyuan | oh really? | 17:05 |
| Ziyuan | I can submit one more patch before Friday I think... | 17:05 |
| @sonney2k | Submitting more patches doesn't really improve anyones chances by now. Nevertheless, we naturally like to see you all contributing and very welcome any patch. Becoming a long term contributor is likely to get you into next years previous years. This is not to say you are (not) in. Official announcements will be made next Monday only. | 17:17 |
| siddharth | sonney2k, btw did u check my pull request | 17:20 |
| @sonney2k | gtg | 17:21 |
| @sonney2k | l8r | 17:21 |
| siddharth | ok see you | 17:22 |
| -!- siddharth [~siddharth@117.211.88.150] has quit [Remote host closed the connection] | 17:26 | |
| -!- siddharth [~siddharth@117.211.88.150] has joined #shogun | 17:28 | |
| Ziyuan | Yep, I understand | 17:46 |
| -!- josip [~josip@unaffiliated/josip] has quit [Ping timeout: 252 seconds] | 18:01 | |
| alesis-novik | Good day | 18:12 |
| @sonney2k | hi... | 18:15 |
| @sonney2k | bettyboo, come on good morning! | 18:15 |
| @bettyboo | sonney2k: yes - it is all good now! | 18:15 |
| * sonney2k giggles | 18:15 | |
| alesis-novik | sonney2k, I removed GMM from current pull request | 18:33 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 18:35 | |
| alesis-novik | sonney2k, I'll close the current one and start one that actually makes sense in the description | 18:42 |
| dvevre | anyone know why AsciiFile.h has: #include <lib/common.h> and not <shogun/lib/...> ? | 18:50 |
| dvevre | and how is it working in that condition? | 18:50 |
| -!- blackburn [~qdrgsm@188.168.4.97] has joined #shogun | 18:54 | |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 18:55 | |
| -!- dvevre [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 18:55 | |
| blackburn | sonney2k: oh, thanks for fixed it | 18:57 |
| -!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Ping timeout: 252 seconds] | 19:00 | |
| alesis-novik | sonney2k, submitted the both separate pull requests | 20:15 |
| -!- josip [~josip@unaffiliated/josip] has joined #shogun | 20:21 | |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 20:31 | |
| -!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 20:36 | |
| -!- Ziyuan [~Ziyuan@116.21.255.137] has quit [] | 20:37 | |
| @sonney2k | blackburn, just remember that include paths should be relative to libshogun/ next time | 20:44 |
| @sonney2k | alesis-novik, thanks | 20:44 |
| -!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 20:50 | |
| CIA-110 | shogun: Ziyuan Lin master * r1be7c44 / src/libshogun/kernel/BesselKernel.h : Copyright infomation revised - http://bit.ly/gFJG62 | 20:51 |
| siddharth | sonney2k, thanks for pointing out the mistakes | 20:51 |
| siddharth | I will correct them accordingly | 20:51 |
| siddharth | as for the testing part, I tested on a small dataset | 20:52 |
| siddharth | so maybe will try on toy dataset perhaps | 20:52 |
| siddharth | or if u have any in mind? | 20:52 |
| josip | sonney2k: what are the main issues with PerformanceMeasures? | 20:53 |
| josip | they seem to follow the standard ROC computational algorithm | 20:53 |
| josip | computation* | 20:53 |
| @sonney2k | josip, blackburn fixed them all now I think | 20:53 |
| josip | oh, ok | 20:54 |
| @sonney2k | josip, it was just not cleanly designed and computing area under precision recall curve needs to be sped up | 20:54 |
| josip | it uses the trap. rule | 20:54 |
| @sonney2k | still only a greedy algorihtm exists | 20:54 |
| josip | hm, okay | 20:55 |
| josip | is there anything else I can do? | 20:55 |
| @sonney2k | alesis-novik, why does get_log_likelihood_example() return 1 ? | 20:56 |
| alesis-novik | sonney2k, sorry , I think this is the old version | 20:57 |
| siddharth | sonney2k, on which dataset did u tested SVMSGD? | 20:57 |
| @sonney2k | josip, I don't currently have ideas for easy tasks.... but as I said previously | 20:59 |
| @sonney2k | Submitting more patches doesn't really improve anyones chances by now. Nevertheless, we naturally like to see you all contributing and very welcome any patch. Becoming a long term contributor is likely to get you into next years previous years. This is not to say you are (not) in. Official announcements will be made next Monday only. | 20:59 |
| alesis-novik | sonney2k, there, it should have been NOTIMPLEMENTED | 20:59 |
| @sonney2k | alesis-novik, yes that would be much preferred | 20:59 |
| alesis-novik | I changed it. I think I just had forgotten to commit it | 20:59 |
| @sonney2k | siddharth, you could use one of the data sets from their paper http://www-etud.iro.umontreal.ca/~bordesa/mywiki/lib/exe/fetch.php?media=bordes_erratumsgdqn.pdf | 21:01 |
| @sonney2k | or the large scale learning challenge | 21:01 |
| siddharth | ok will try on that | 21:01 |
| @sonney2k | http://largescale.ml.tu-berlin.de/instructions/ | 21:01 |
| @sonney2k | siddharth, he actually submitted sgd-qn there http://largescale.ml.tu-berlin.de/submission/evaluation/1/5/ | 21:02 |
| -!- josip [~josip@unaffiliated/josip] has quit [Ping timeout: 246 seconds] | 21:03 | |
| siddharth | Newton SVM rank 1...great | 21:04 |
| @sonney2k | siddharth, newton svm and sgd-qn were the fastest... depends on which machine you evaluate / data ... | 21:05 |
| siddharth | ohh I see | 21:05 |
| -!- josip [~josip@95.180.215.37] has joined #shogun | 21:06 | |
| -!- josip [~josip@95.180.215.37] has quit [Changing host] | 21:07 | |
| -!- josip [~josip@unaffiliated/josip] has joined #shogun | 21:07 | |
| josip | sonney2k: okay | 21:07 |
| siddharth | sonney2k, what name do you suggest for the 'matrix mul' function? | 21:11 |
| siddharth | basically its a diagnal matrix represented in vector form | 21:12 |
| @sonney2k | siddharth, I don't have a good name - it is some componentwise multiplication of a vector - sth like component_wise_vector_multiply() - I am open for better names | 21:13 |
| * sonney2k only 2 more days and my mobile phone's data plan will be back to full speed :) | 21:14 | |
| siddharth | will think of a smaller name perhaps... | 21:15 |
| josip | diag_multiply ? | 21:15 |
| josip | :D | 21:15 |
| @sonney2k | siddharth, why not vector_multiply? | 21:15 |
| siddharth | yeah thats good | 21:16 |
| @sonney2k | dot() is already there for dot product | 21:16 |
| @sonney2k | so no confusion possible | 21:16 |
| siddharth | great will name vector_multiply | 21:16 |
| josip | dot product results in a scalar, this is not the same - right? | 21:16 |
| siddharth | yeah | 21:17 |
| siddharth | output is a vector | 21:17 |
| @sonney2k | josip, no each element is multiplied with a component from some other vector | 21:17 |
| josip | exactly | 21:18 |
| * blackburn vernulsya | 21:55 | |
| blackburn | sonney2k: I really wonder why I didn't think about unproper includes.. | 21:56 |
| blackburn | next time will know :) | 21:56 |
| @sonney2k | blackburn, heh | 22:05 |
| @sonney2k | now I hope that you get well soon | 22:05 |
| blackburn | what you mean? | 22:05 |
| @sonney2k | blackburn, I thought you were catching a cold? | 22:05 |
| blackburn | aha, sitting there and sneezing | 22:06 |
| blackburn | ah, that's what you mean ;) | 22:06 |
| blackburn | thank you | 22:06 |
| alesis-novik | Spring is the worst time, I never get sick during winter, only spring | 22:06 |
| @sonney2k | I only have 2 more weeks left - then we will be getting a new baby.... | 22:07 |
| * sonney2k prepares for 0 sleep | 22:07 | |
| blackburn | sonney2k: oh! congrats to you! | 22:07 |
| alesis-novik | congratulations sonney2k | 22:07 |
| blackburn | are you already have one? | 22:07 |
| @sonney2k | yes | 22:07 |
| @sonney2k | small one still | 22:07 |
| @sonney2k | just learned to walk... | 22:07 |
| blackburn | what is his name? :) | 22:08 |
| @sonney2k | not yet very stable though | 22:08 |
| josip | congrats | 22:08 |
| * blackburn hopes for not getting one in next 5 years :D | 22:08 | |
| @sonney2k | and she is pretty scared of 'ants' ... tries hard to say the german word. 'Ameisen' | 22:08 |
| blackburn | she.. sorry | 22:08 |
| @sonney2k | blackburn, that will turn you into a productive beast | 22:08 |
| @sonney2k | blackburn, Livia | 22:09 |
| blackburn | nice :) | 22:09 |
| @bettyboo | :> | 22:09 |
| @sonney2k | next one I will tell in 2-3 weeks :) | 22:09 |
| * sonney2k hopes that blackburn can take over while he is busy with childcare | 22:09 | |
| blackburn | sonney2k: take over what? | 22:10 |
| josip | reminds me of Paul Graham's comment on how to become a morning person "Getting lots of exercise makes it easier to fall asleep at night. The ultimate solution, however, is to have a small child." | 22:10 |
| blackburn | sonney2k: shogun? :) how it could be? | 22:12 |
| @bettyboo | ;D | 22:12 |
| alesis-novik | sonney2k, I take it the GMM will only be merged in when it's in the final form after the summer? | 22:12 |
| @sonney2k | blackburn, shogun of course - you already have 72 patches - turns you into rank #7 in just 2 weeks. | 22:15 |
| @sonney2k | alesis-novik, I would merge whenever it compiles + sth new works | 22:16 |
| blackburn | some of my patches are 'Commit the Great' removing '_' from 'k_' :D | 22:16 |
| @sonney2k | I prefer small changes over completely ready ones | 22:16 |
| alesis-novik | sonney2k, so the current one was just too simple? | 22:17 |
| @sonney2k | blackburn, speaking of which - there is still the rename CClassifier -> CMachine pending | 22:17 |
| blackburn | sonney2k: eh, it is just renaming issue? | 22:17 |
| @sonney2k | alesis-novik, no - what I meant to say is that pull requests that do one particular thing are much easier to read for me | 22:18 |
| @sonney2k | alesis-novik, it is very difficult to review patches and don't miss a thing | 22:18 |
| blackburn | sonney2k: this week I think I will work on ROC etc and some work on CGraph | 22:18 |
| @sonney2k | and currently the test suite still does not run through... | 22:18 |
| @sonney2k | blackburn, the auROC algorithm should be fine ... auPRC is not yet there | 22:19 |
| blackburn | which tests fails? | 22:20 |
| @sonney2k | blackburn, yes I thiink the confusion of CClassifier & classify can only be resolved by calling things apply | 22:20 |
| @sonney2k | blackburn, run make tests and you will see ... plenty | 22:20 |
| alesis-novik | sonney2k, I was just asking why you closed the GMM pull request without any comment :) | 22:20 |
| @sonney2k | alesis-novik, I thought that was the wrong one? | 22:21 |
| @sonney2k | at least you said that there shoudl be SG_NOTIMPLEMENTED in likelihood or derivative tehre? | 22:21 |
| alesis-novik | sonney2k, yeah, and I pushed the changes on it :D | 22:21 |
| @sonney2k | alesis-novik, well then send another request: btw I also noted that you don't have newlines after and before loops ... that is also sth I would like to have for readability | 22:22 |
| blackburn | SubGradientSVM, HistogramWordString, LocalAlignmentString,... heheheh | 22:22 |
| @sonney2k | SubgradientSVM - I don't care but the rest should have worked.... | 22:23 |
| alesis-novik | sonney2k, I added that as requested in the previous pull as well | 22:23 |
| @sonney2k | alesis-novik, wasn't there I am pretty sure | 22:24 |
| * sonney2k checks | 22:24 | |
| @sonney2k | alesis-novik, https://github.com/shogun-toolbox/shogun/pull/56/files also removes one newline after a for loop | 22:25 |
| @sonney2k | (line 57 that is | 22:25 |
| alesis-novik | sonney2k, so I should add that line back | 22:26 |
| @sonney2k | yes please | 22:27 |
| @sonney2k | otherwise it can be misread to belong to the for loop | 22:27 |
| @sonney2k | alesis-novik, regarding the other patch I have 2 very minor comments too - will add them | 22:30 |
| alesis-novik | added the newline, awaiting comments | 22:30 |
| blackburn | why there are about 5 new forks today?.. | 22:30 |
| blackburn | it seems to be a bit late for rushing for contibuting :) | 22:31 |
| * blackburn remembered where he saw an Alesis' last name | 22:32 | |
| @sonney2k | alesis-novik, comments sent | 22:33 |
| @sonney2k | blackburn, no idea | 22:33 |
| * sonney2k hopes that some people just keep on posting patches even though not being part of GSoC later | 22:33 | |
| @sonney2k | blackburn, btw are you ok when I rename CClassifier to CMachine? | 22:34 |
| @sonney2k | or should it be CMethod? | 22:34 |
| @sonney2k | Machine sounds cooler though | 22:34 |
| blackburn | sonney2k: I have absolutely no idea (and opinion), just do it ;) | 22:34 |
| @sonney2k | blackburn, does kernelmachine sound good to you? | 22:35 |
| blackburn | yeap | 22:35 |
| @sonney2k | blackburn, recall the discussion about class hierarchy and that regression methods were derived from classifiers? | 22:35 |
| @sonney2k | my proposed fix is to rename it to CMachine and its classify function to apply() | 22:36 |
| blackburn | a little.. so with renaming we could solve the problem of being regression algos CClassifier(s), right? | 22:36 |
| @sonney2k | at least they are then derived from CMachine - much more general name | 22:36 |
| alesis-novik | sonney2k, submitted changes | 22:37 |
| blackburn | yeap, sound good | 22:37 |
| @sonney2k | blackburn, ok then I will do this rename *when the testsuite* fully works again | 22:37 |
| blackburn | sonney2k: даю добро, but don't know why my opinion is important ;) | 22:38 |
| @sonney2k | alesis-novik, that code now is a pure beauty to look at! | 22:39 |
| CIA-110 | shogun: Soeren Sonnenburg master * r6d779fb / (2 files): Merge branch 'gmm' of https://github.com/alesis/shogun (+9 more commits...) - http://bit.ly/f3qlOZ | 22:39 |
| * blackburn is interested now to see a beauty | 22:40 | |
| @sonney2k | blackburn, dobro was ~ good? but daju ? | 22:40 |
| blackburn | sonney2k: I don't pretty sure if english have this phrase, daju is giving | 22:41 |
| josip | добро = good in macedonian as well | 22:41 |
| josip | same in bulgarian and serbo-croatian :D | 22:41 |
| @bettyboo | ;D josip | 22:41 |
| blackburn | josip: why do you know it? ;) | 22:41 |
| josip | well macedonian is my native language and I can understand serbo-croatian and bulgarian | 22:42 |
| blackburn | josip: ah, I see | 22:43 |
| blackburn | don't know anything in srpska | 22:43 |
| blackburn | except kosovo je srbija :D | 22:43 |
| @sonney2k | blackburn, you only need 31 times more patches to beat me - will take you ~1.5 yrs at your curent pace | 22:44 |
| josip | lol :)) | 22:45 |
| josip | blackburn: you're from ? | 22:45 |
| blackburn | josip: russia | 22:45 |
| josip | oh I see | 22:45 |
| blackburn | sonney2k: i'll try ;) | 22:45 |
| josip | you're a newcomer to the project as well? | 22:45 |
| blackburn | josip: yeap, applying for dim.reduction | 22:45 |
| @sonney2k | blackburn, but please don't beat me just in this summer | 22:46 |
| @sonney2k | alesis-novik, thanks applied | 22:46 |
| * sonney2k is tired again | 22:46 | |
| @sonney2k | good night everyone | 22:46 |
| blackburn | sonney2k: I will have exams in june so I have to spent a little less time for shogun | 22:46 |
| blackburn | sonney2k: good night | 22:46 |
| @sonney2k | l8r | 22:46 |
| josip | good night | 22:47 |
| alesis-novik | night sonney2k | 22:47 |
| blackburn | bettyboo: will you marry alesis-novik? | 22:47 |
| @bettyboo | blackburn: then you should get some more meaningful error msg | 22:47 |
| alesis-novik | I don't even know how to interpret that :D | 22:48 |
| blackburn | bettyboo: why do you talking about error msg, will you marry? you are 42yr old virgin! | 22:48 |
| @bettyboo | blackburn: hello my network had some problem just now | 22:48 |
| josip | lol | 22:49 |
| alesis-novik | :D | 22:49 |
| blackburn | bettyboo: so, marry or not? | 22:49 |
| @bettyboo | blackburn: why not? | 22:49 |
| blackburn | alesis-novik: i will be тамада | 22:49 |
| blackburn | btw, how it will be in english? really don't know | 22:50 |
| alesis-novik | I'm not even sure what it means in russian | 22:51 |
| blackburn | :D | 22:51 |
| blackburn | master of ceremonies, guy who leading marriage celebrates, etc | 22:52 |
| alesis-novik | Ah | 22:53 |
| alesis-novik | So, with all slots assigned, it's just the matter of waiting this one week | 23:07 |
| blackburn | alesis-novik: Soeren said it? | 23:07 |
| alesis-novik | <sonney2k> Submitting more patches doesn't really improve anyones chances by now. Nevertheless, we naturally like to see you all contributing and very welcome any patch. Becoming a long term contributor is likely to get you into next years previous years. This is not to say you are (not) in. Official announcements will be made next Monday only. | 23:10 |
| alesis-novik | I'll see you later folks :) | 23:28 |
| -!- alesis-novik [~alesis@188.74.87.84] has quit [Quit: I'll be Bach] | 23:29 | |
| --- Log closed Tue Apr 19 00:00:36 2011 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!