Open in new window / Try shogun cloud
--- Log opened Sat Aug 06 00:00:09 2011
@sonney2kblackburn, do you know what is?00:08
CIA-87shogun: Soeren Sonnenburg master * r7013309 / (2 files in 2 dirs):00:09
blackburnsonney2k: it is byte output stream able to write/read objects00:09
CIA-87shogun: derive java modular classes from Serializable and add dummy00:09
CIA-87shogun: write/readObject functions -
CIA-87shogun: Soeren Sonnenburg master * rf27d051 / (40 files in 9 dirs): Merge branch 'master' of -
@sonney2kblackburn, because I managed to do the changes00:09
@sonney2kto java modular00:09
@sonney2kthe problem now is how to use shogun's self.save_serializable() with these streams00:10
blackburnyou probably should pull before commit hehe00:10
blackburnsonney2k: OutputStream is a base class for output (write) byte streams00:11
blackburnit is able to write bytes, not some chars00:11
blackburnany *OutputStream is an facade or so00:11
blackburne.g. ObjectOutputStream to write objects00:12
@sonney2kso chars would work as well - currently I need to create some temporary file with this on disc - then load this and then I could stream the stuff java stylequestion is whether I should create a temporary file or00:13
@sonney2kthis would involve some new io thingy that does everything in memory00:13
blackburnhmm why you want to make temporary file?00:13
-!- serialhex [] has quit [Read error: Connection reset by peer]00:14
@sonney2kblackburn, to dump the object in there - it is all filename based currently00:14
blackburnif you want to use chars you could use OutputStreamWriter class00:14
blackburnit is designed to be adapter between char and byte streams00:15
blackburnsonney2k: so when you want to serialize object it is stored in file and after some interface load it?00:17
@sonney2kblackburn, in shogun you can just write somesgobject.save_serializable(someserializablefile)00:23
blackburnhow could it work with java?00:23
@sonney2ksomeserializablefile can be xml dump of shogun objects, ascii file, hdf5, json ...00:23
@sonney2kI don't understand the question00:24
@sonney2khow not?00:24
@sonney2kif one implements serializable interface00:24
blackburnof course it will00:24
blackburnI mean how could you use it00:24
blackburnin java write/readObject00:24
@sonney2kand then the writeObject/readObject functions (which I would make to work by just doing sgobject.save_serializable(serializableasciifile('/tmp/x') then load that with java and write in this out stream)00:25
@sonney2ksame with loading00:25
blackburnokay, got it00:26
@sonney2kI only meant one could save i/o if one does this immediately in memory avoiding the tmp file00:26
blackburnyes, I would say it is better00:27
@sonney2kyeah but not implemented right now00:27
blackburnI see00:27
@sonney2kso I will first do the file bases solution and then we can improve later (no user visible change...)00:28
blackburnI'll make a try to make distances parallel now00:29
@sonney2kI need sleep00:30
blackburnsee you00:30
-!- serialhex [] has joined #shogun00:31
CIA-87shogun: Sergey Lisitsyn master * r86e9ed9 / (2 files): Made CDistance parallel -
blackburnokay sleep now02:09
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]02:12
-!- blackburn [~blackburn@] has joined #shogun12:06
-!- f-x [~user@] has joined #shogun12:10
f-xsonney2k: hey, are you free?12:21
-!- f-x [~user@] has quit [Remote host closed the connection]12:27
-!- f-x [~user@] has joined #shogun12:37
-!- f-x [~user@] has quit [Ping timeout: 260 seconds]13:35
-!- f-x [~user@] has joined #shogun14:00
-!- blackburn [~blackburn@] has quit [Ping timeout: 255 seconds]14:03
-!- blackburn [~blackburn@] has joined #shogun14:05
CIA-87shogun: Sergey Lisitsyn master * r74f775f / (2 files): Made Isomap Heap-Dijkstra parallel -
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]15:08
-!- blackburn [~blackburn@] has joined #shogun15:08
blackburnSG_FREE(sonney2k) :D15:09
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]15:40
-!- heiko [~heiko@] has joined #shogun16:28
-!- heiko [~heiko@] has quit [Remote host closed the connection]16:29
-!- heiko [~heiko@] has joined #shogun16:35
heikosonney2k, it is not possible to use templates with two parameters right? class_list.cpp instanciates always with one type16:36
-!- heiko [~heiko@] has quit [Quit: Leaving.]17:37
-!- f-x [~user@] has quit [Ping timeout: 260 seconds]17:55
-!- f-x [~user@] has joined #shogun19:14
-!- blackburn [~blackburn@] has joined #shogun19:48
f-xsonney2k: could we make index_t unsigned int? (it's signed int now) and also including the typedef in common.h would be very convenient21:52
@sonney2kf-x, making it unsigned int will probably cause a couple of issues22:06
@sonney2kf-x, so it is timing wise a bad idea when we want a release in 3-4 weeks.22:06
@sonney2kbut it certainly should be in common.h22:06
CIA-87shogun: Alesis Novik master * r94d79dc / src/shogun/clustering/GMM.cpp : Some fixes and checking for illegal use. -
CIA-87shogun: Soeren Sonnenburg master * re1eac74 / src/shogun/clustering/GMM.cpp :22:07
CIA-87shogun: Merge pull request #269 from alesis/gmm22:07
CIA-87shogun: Some fixes and checking for illegal use. -
f-xthen is it ok to use size_t? because replacing with index_t would cause some incompatibilites (that could be dealt with if i replace size_t with uint32_t but i don't like that)22:07
@sonney2kf-x, size_t is a dangerous choice22:08
@sonney2kbecause you don't know if it is 32 or 64 bit22:08
@sonney2kf-x, I am just reading your patch - why didn't you move substring to SGIO ?22:09
f-xsonney2k: because i would need to rewrite the IO functions to use it22:10
@sonney2kI mean these are only isolated helper functions22:10
f-xsonney2k: you mean i should move substring to SGIO but not do the get_vector_* modifications now?22:13
f-xwe should make AsciiFile buffered later on too, but it seems using the buffer is slower on my computer, and faster on yours and John's22:16
f-xso i want to wait for some more testing22:16
@sonney2kf-x, but it was not really slower... I mean we were talking 1 mu sec or so22:16
f-xsonney2k: that's right, but again the improvement also is just 1 mu sec, so I was kind of hesitant to replace a thing like AsciiFile22:18
@sonney2kf-x, yeah but you compared agains GNU getline...22:18
f-xanyway i have the buffered readers more or less ready22:19
f-xsonney2k: yes22:19
f-xokay, so it would make a huge difference on OS X22:19
f-xfor which the flimsy version is implemented22:19
@sonney2kf-x, anyway - so move it to SGIO22:20
@sonney2kand then we will see :)22:20
f-xsonney2k: is it better if i make substring a class and include the functions as member functions?22:28
f-xor is it ok the way it is?22:28
f-xthe only difference would be s.float_of_substring() instead of float_of_substring(s)22:28
@sonney2kf-x, no don't make it a class22:34
@sonney2kserialhex, around?22:37
@sonney2kserialhex, if so time to rub-rub-ruby?22:37
f-xsonney2k: is including <string> in SGIO allowed? or should it be avoided?22:49
@sonney2kwould be great if you could avoid it22:50
@sonney2k(like all std:: stuff)22:50
@sonney2kf-x, blackburn sry but I need to sleep *again*23:18
blackburnsee you :)23:19
f-xsonney2k: see ya23:22
-!- f-x [~user@] has quit [Remote host closed the connection]23:37
--- Log closed Sun Aug 07 00:00:20 2011