Open in new window / Try shogun cloud
--- Log opened Tue Aug 15 00:00:39 2017
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has quit [Quit: Page closed]00:34
-!- Netsplit *.net <-> *.split quits: micmn01:54
-!- Netsplit *.net <-> *.split quits: sonney2k, shogitter, zxtx, OXPHOS, ironstark, @ChanServ01:54
-!- Netsplit over, joins: micmn01:57
-!- Netsplit over, joins: OXPHOS, @ChanServ01:58
-!- ironstark [sid221607@gateway/web/] has joined #shogun01:58
-!- ironstark [sid221607@gateway/web/] has quit [Changing host]01:58
-!- ironstark [sid221607@gateway/web/] has joined #shogun01:58
-!- Netsplit over, joins: shogitter, sonney2k, zxtx01:59
-!- Netsplit *.net <-> *.split quits: OXPHOS, @ChanServ01:59
-!- Netsplit *.net <-> *.split quits: micmn02:00
-!- Netsplit over, joins: micmn02:02
-!- Netsplit over, joins: OXPHOS, @ChanServ02:02
@sukey[] Pull Request opened by Sahil33307:27
-!- geektoni [] has joined #shogun09:48
@sukey[] Pull Request synchronized by Sahil33310:12
@wikinggeektoni, ping?11:11
geektoniwiking: pong11:11
@wikinganything blocking you/11:11
@wikingyesterday i've passed out11:11
@wikingdunno if there's anything that i should help u with11:12
@wikingor not11:12
geektoniwiking: well, there is still the python interface thing11:12
geektoniso, yesterday I was able to make the get_observations() object iterable11:13
geektoniI mean that this works now11:15
geektoniwiking: however, another little problem arose here
@wikingaah you mean the casting :)11:18
geektonisince I store the trained machines as CMachine*, I have to cast them to the proper type when I need to use them11:18
geektoniyes yes11:18
@wikingso afaik there's no obtain_from_generic :P11:18
@wiking(see labels has that operator)11:18
geektoniok, so I'll need to add it11:22
@wikingyep :D11:22
geektoniwiking: is there a general rule where to place obtain_from_generic methods? because I see they're implemented here and there :P11:24
@wikingno it's crazy :)11:24
@wikingit was on demand basis11:25
geektoniI'm not surprised ;)11:28
@wiking:D well enterprise design11:29
-!- HeikoS [] has joined #shogun11:48
-!- mode/#shogun [+o HeikoS] by ChanServ11:48
@wikingHeikoS, ping?12:08
@HeikoSwiking: jo12:08
@wikingso the only question now is here12:09
@sukey[] Pull Request synchronized by Sahil33312:11
@wikingbtw have u seen this ^ PR?12:12
@HeikoSwiking: didnt see it yet12:16
@HeikoSwill check12:16
@HeikoSwiking: yeah we can discuss this with the team12:16
@HeikoShave a voting system or so12:16
@HeikoSbut I think two things make sense: zurich or hungatry12:16
@wikingi've picked up the thread12:18
@wikingon the mailinglist12:18
@sukey[] Pull Request
@sukey[] Pull Request synchronized by Sahil33313:02
-!- sheogorath27 [sheogorath@gateway/shell/firrre/x-rdgcmdvvjpibutry] has left #shogun []13:27
@sukey[] Pull Request synchronized by MikeLing14:11
-!- HeikoS [] has quit [Ping timeout: 240 seconds]14:31
@sukey[] Pull Request synchronized by geektoni15:58
geektoniwiking: ^ I've fixed the last things. It should be ready now (let's see what CIs say).16:06
-!- olinguyen [81615ad9@gateway/web/freenode/ip.] has joined #shogun16:30
geektoniwiking: is there any way to make the meta example generator produce c++ class instantiations without the "C" prefix? :P without touching the generator or the cpp.json file I mean :P16:47
geektoniwiking: btw, is all green16:49
-!- HeikoS [] has joined #shogun17:04
-!- mode/#shogun [+o HeikoS] by ChanServ17:04
olinguyenHeikoS: Hey, wanna chat?17:05
-!- HeikoS [] has quit [Ping timeout: 248 seconds]17:08
-!- OXPHOS_ [92bd305b@gateway/web/freenode/ip.] has joined #shogun17:15
-!- OXPHOS_ [92bd305b@gateway/web/freenode/ip.] has quit [Client Quit]17:16
-!- HeikoS [] has joined #shogun17:23
-!- mode/#shogun [+o HeikoS] by ChanServ17:23
olinguyenHeikoS: hey17:25
@HeikoSolinguyen:  hi!17:25
olinguyendo you have some time to chat a little?17:25
@HeikoSchat is OK17:26
@HeikoSI cannot do a call, as I have some stuff to do, will have to do that tomorrow17:26
olinguyenthat's fine17:26
olinguyenSo about SQL vs Pandas: i found that it would end up with more computation with pandas17:27
olinguyencause for a single patient and single variable (heart rate) it amounts to ~1,600,000 data points17:27
olinguyenall patients*17:27
olinguyenbut for only heart rate17:28
olinguyenexpanding that to other vital signs, i'd have to do a separate query each time17:28
olinguyenand then resample and combine through pandas17:28
@HeikoSolinguyen: I see17:29
@HeikoSso with SQL you never even touch the data you dont need17:29
olinguyenexactly, and i can aggregate all vital signs within that hour timestamp17:29
@HeikoSthat is fine then17:29
@HeikoSmake sure to do some (reproducible) testing that pandas and SQL re-sampling gives you the same result17:29
@HeikoS(on small time windows of course)17:29
olinguyenyea, i briefly double checked. it looks consistent17:30
olinguyeni think you had other questions. do you want to leave that for the call tomorrow or so?17:31
@HeikoSolinguyen: yeah17:32
@HeikoSolinguyen: so were the thing sin the plots for one patient or for all?17:33
olinguyena single patient17:33
@HeikoSI guess for all of them17:33
@HeikoSok and the evaluation as well only for one?17:33
olinguyenevaluation is for all patients17:33
@HeikoSI see17:33
@HeikoSok thats good then17:33
@HeikoSyou should try to experiment with different target labels17:34
olinguyeni split the train/test to 50/5017:34
@HeikoS"dies within the next week" "dies within next month"17:34
@HeikoSnote that those might include each other17:34
@HeikoSif you use "within"17:34
@HeikoSbut "within" is better than "at"17:34
@HeikoS50/50 of all available data?17:34
@HeikoSand whats the class ratio for the labels you used?17:34
olinguyeni didn't do any undersamping (left it as is)17:35
olinguyeni'll see how changing the ratio affects the results17:35
@HeikoSin theory, you should only have a single "positive" point per patient at the moment no?17:36
@HeikoSwhat exactly was your label?17:36
olinguyenit depends where in time we're looking at17:36
olinguyenstarting at 00:00, if the patient dies at 23:00, and we look at the sample at say 8:00, then the label is alive in 6 hours17:37
olinguyenchanging to "dies in 1 day" would be different17:37
olinguyenbut it seems that the classifier score is pretty similar for both outcomes (dies in 6 hours or dies in 1 day)17:38
@HeikoSolinguyen: what does your "in" mean?17:44
@HeikoSor at?17:44
olinguyenevery row, or sample has a timestamp. Looking whether the patient is alive/dead in 6 hours means i'm looking at 6 hours ahead of time, from the time of that timestamp17:46
@HeikoSolinguyen: ok, so that is "within"17:46
@HeikoSsince a patient never is alive after once being dead17:47
@HeikoSso you get multiple positives for a single death event17:47
olinguyenthat's correct17:47
@HeikoSthat is good17:47
@HeikoSmakes the problem slightly easier17:47
@HeikoScan you try multiple time horizonts?17:47
@HeikoSlike a day, a week, a month17:47
@HeikoSa year?17:47
@HeikoSand the same for the features17:48
olinguyenyea, but i'd have to base it on a few days of vital signs17:48
olinguyenbecause the majority of patients stay in the ICU for < 5 days17:48
@HeikoScomputing averaged for 6h, 1d, 1w, 1y17:48
@HeikoSah I see17:48
@HeikoSolinguyen: ok let's talk tomorrow17:48
@HeikoSi have to finish some stuff here17:49
OXPHOSJust came back from seattle :)17:53
OXPHOSThere's a cppcon by the end of Sept there17:53
OXPHOSthinking about going17:53
-!- mikeling [uid89706@gateway/web/] has joined #shogun18:24
mikelingHeikoS: Hi, does gpl module is necessary to build every time?  Could we avoid to build it by run 'make'?18:25
@HeikoSmikeling: you can turn int off18:25
@HeikoSOXPHOS: hello!18:25
mikelingoh! how18:26
@HeikoScmake option18:26
mikelingok thank you@18:26
@HeikoSlet me know if it worked18:26
mikelingit looks like I should turn off the SVMLight  also?18:31
mikelingit says Can only use SVMLight when GPL codes are included18:31
@wikingHeikoS, i just remember that i wanted to ask... what happened with src/shogun/statistical_testing/internals/FeaturesUtil.cpp ?18:35
@wikingit is dropped for good?18:35
@HeikoSmikeling: yes SVMLight has GPL dependencies18:42
@HeikoSwiking: I don't know18:42
@wikingHeikoS, just because it's not in gpl18:43
@wikingbut used to be in shogun18:43
@wikinghence i'm just wondering what happened18:43
mikelingHeikoS: yes, I just turn it off and everything works :)18:43
@HeikoSI don tknow18:43
@HeikoSdont remember changing things18:44
@HeikoSbut maybe rahul did?18:44
-!- geektoni [] has quit [Remote host closed the connection]19:20
@sukey[] Pull Request synchronized by MikeLing19:31
-!- HeikoS [] has quit [Ping timeout: 255 seconds]19:35
-!- Netsplit *.net <-> *.split quits: mikeling21:05
-!- mikeling [uid89706@gateway/web/] has joined #shogun21:06
--- Log closed Wed Aug 16 00:00:41 2017