--- Log opened Wed Apr 27 00:00:37 2011 | ||
blackburn | sonney2k: finished work on shogun.Evaluation, sent a pull request | 00:27 |
---|---|---|
-!- PD [53174b7b@gateway/web/freenode/ip.83.23.75.123] has joined #shogun | 00:46 | |
-!- PD is now known as Guest99107 | 00:47 | |
Guest99107 | Hi everyone. Could anyone please tell me if I can define in shogun my own distance for kmeans? | 00:50 |
blackburn | Guest99107: hi. what interface do you use? | 00:56 |
blackburn | I mean python or ..? | 00:56 |
blackburn | Guest99107: anyway, just take a look at shogun/examples/undocumented/*your interface*/clustering_kmeans* | 00:58 |
blackburn | if I understood you right and you do not want to define your own distance not implemented in shogun, cause there is no way to do it | 00:59 |
Guest99107 | I can use any | 01:13 |
Guest99107 | does not matter | 01:13 |
Guest99107 | just would like to define my own distance for kmeans | 01:13 |
Guest99107 | and I wondering if it is possible | 01:14 |
Guest99107 | *I am | 01:14 |
Guest99107 | ? | 01:15 |
blackburn | then just take a look at /examples/undocumented/python_modular/clustering_kmeans_modular.py | 01:18 |
blackburn | there are EuclidianDistance and you can easily change it to any other available | 01:18 |
blackburn | hope I helped you, have to go sleep now | 01:23 |
-!- blackburn [~qdrgsm@188.168.5.125] has quit [Quit: Leaving.] | 01:23 | |
Guest99107 | I do not see a way to modify the distance | 01:25 |
Guest99107 | I think it is just a wrapper for c/c++ class | 01:25 |
-!- Guest99107 [53174b7b@gateway/web/freenode/ip.83.23.75.123] has quit [Quit: Page closed] | 02:34 | |
-!- alesis-novik [~alesis@188.74.87.84] has joined #shogun | 02:34 | |
sploving | hello blackburn | 03:23 |
-!- sploving [~root@124.16.139.196] has quit [Read error: Operation timed out] | 04:19 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 04:20 | |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 04:55 | |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 07:21 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 07:23 | |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Quit: Page closed] | 07:31 | |
-!- akhil_ [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 07:36 | |
@sonney2k | sploving, around now? | 07:39 |
sploving | yep. morning | 07:39 |
@bettyboo | you here?! | 07:39 |
sploving | sonney2k, so early. it is not 7:00 o'clock in your timezone | 07:41 |
@sonney2k | sploving, which time zone are you in | 07:41 |
sploving | utc+8 | 07:42 |
sploving | now it is 13:42 | 07:42 |
@sonney2k | I usually get up before 7 am - I have a kid already :) | 07:42 |
sploving | i tried to test your example | 07:42 |
sploving | but failed | 07:42 |
@sonney2k | sploving, but you could compile w/ java correctly? | 07:43 |
@sonney2k | I mean shogun first of all | 07:43 |
@sonney2k | sploving, with ./configure --interfaces=libshogun,java_modular | 07:43 |
sploving | I could compile it (using git pull) | 07:43 |
@sonney2k | and then cd java_modular | 07:44 |
sploving | yeap | 07:44 |
@sonney2k | java *.java path_to_example.java | 07:44 |
@sonney2k | ^^ too? | 07:44 |
@sonney2k | err javac | 07:44 |
@sonney2k | javac *.java path_to_example.java | 07:44 |
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | We participate in GSoC with 5 projects http://bit.ly/eI31bL | Kick-off meeting Wednesday Apr 27 at 13:00 UTC | Channel logs http://bit.ly/eQDk8Q | 07:46 | |
sploving | sonney2k, you mean javac *.java examples/undocumented/java_modular | 07:46 |
@sonney2k | sploving, well the filename of the example | 07:46 |
sploving | sonney2k, you mean javac *.java examples/undocumented/java_modular/shogun.java | 07:46 |
sploving | just an example | 07:47 |
@sonney2k | yes and then LD_LIBRARY_PATH=../libshogun:. java shogun | 07:47 |
@sonney2k | and shogun here also just an example | 07:47 |
@sonney2k | sploving, btw will you have time to attend the meeting at 13:00 UTC today? | 07:47 |
sploving | yeap | 07:48 |
sploving | I will be around. It is 21:00 | 07:48 |
sploving | I will be there | 07:48 |
sploving | on time | 07:48 |
@sonney2k | ok great | 07:50 |
@sonney2k | so back to the java example | 07:50 |
@sonney2k | does it work now? | 07:50 |
sploving | not yet. | 07:51 |
sploving | javac *.java ../../examples/undocumented/minimal_test.java | 07:51 |
sploving | it shows the following error: | 07:51 |
sploving | KernelMachine.java: 14 could not find the symbole | 07:52 |
@sonney2k | sploving, fixed | 07:53 |
CIA-90 | shogun: Soeren Sonnenburg master * rb318a31 / examples/undocumented/java_modular/minimal_test.java : fix name mismatch - http://bit.ly/esS72C | 07:53 |
@sonney2k | try again | 07:53 |
@sonney2k | gtg now | 07:53 |
@sonney2k | l8r | 07:53 |
sploving | bye | 07:53 |
sploving | I will have a try again | 07:53 |
@sonney2k | sploving, did you get it to work now? | 08:37 |
-!- blackburn [~qdrgsm@188.168.5.38] has joined #shogun | 08:38 | |
@sonney2k | it works here so... | 08:40 |
@sonney2k | again: | 08:41 |
@sonney2k | cd java_modular | 08:41 |
@sonney2k | javac *.java ../../examples/undocumented/java_modular/minimal_test.java | 08:41 |
@sonney2k | CLASSPATH=../../examples/undocumented/java_modular:. LD_LIBRARY_PATH=../libshogun:. java minimal_test | 08:41 |
sploving | sonney2k, I did not set the classpath | 08:41 |
@sonney2k | then try the line... it should print out 1.0 | 08:42 |
@sonney2k | blackburn, morning! | 08:42 |
blackburn | sonney2k: morgen! I just woke up at do all the things sloow.. :) | 08:42 |
blackburn | hmm link to my graph in description of pull request is broken | 08:44 |
blackburn | fixed | 08:44 |
@sonney2k | sploving, so does it work now? | 08:46 |
sploving | take a minitu | 08:46 |
sploving | in my machine, it is like:"export CLASSPATH=../../examples/undocumented/java_modular:." | 08:47 |
sploving | i have set the path, but when use the"java minimal_test" | 08:53 |
sploving | it still show errors | 08:53 |
sploving | are you sure, "LD_LIBRARY_PATH=../libshogun:." is OK? | 08:55 |
sploving | sonney2k, it still could not find Kernel. "no Kernel in java.library.path" | 08:59 |
@sonney2k | sploving, it has to be in classpath too | 09:00 |
@sonney2k | that is why '.' | 09:00 |
@sonney2k | do you have a Kernel.so in java_modular ? | 09:00 |
* sonney2k is off for breakfast | 09:00 | |
@sonney2k | l8r | 09:00 |
@sonney2k | sploving, and? | 09:15 |
sploving | i use the java-1.6.0-openjdk | 09:15 |
sploving | so I need set the /etc/profile | 09:15 |
sploving | to make it work well | 09:15 |
sploving | I use your way, but failed | 09:16 |
@sonney2k | you can put it on the cmdline - it should work too | 09:16 |
@sonney2k | ok | 09:16 |
@sonney2k | so it works when you do export etc? | 09:16 |
sploving | not worked not | 09:16 |
sploving | now | 09:16 |
@sonney2k | ? | 09:17 |
@sonney2k | what is the error? | 09:17 |
sploving | I have the Kernel.so, but it still show the error | 09:17 |
sploving | no Kernel in java.library.path | 09:17 |
sploving | i set the path what you told me | 09:18 |
@sonney2k | and you are in java_modular dir? | 09:18 |
sploving | yep | 09:18 |
blackburn | * the sorrow of java modular: part 2 * | 09:18 |
@sonney2k | CLASSPATH=../../examples/undocumented/java_modular:. ? | 09:18 |
@sonney2k | do you have the :. at the end? | 09:18 |
sploving | yeap | 09:19 |
@sonney2k | LD_LIBRARY_PATH=../libshogun:. | 09:19 |
sploving | and also for LD | 09:19 |
@sonney2k | also the . | 09:19 |
sploving | yeap | 09:19 |
sploving | I installed java this morning. | 09:19 |
sploving | just use the apt-get install openjdk-6-jd | 09:20 |
sploving | jdk | 09:20 |
sploving | but not set the classpath and path | 09:20 |
@sonney2k | you could try sun-java6-jdk | 09:20 |
sploving | ok I will have a try | 09:20 |
* blackburn thinks doesn't matter: sun or open | 09:20 | |
blackburn | for such example I mean | 09:21 |
blackburn | sonney2k: in russian precision and accuracy is a same word :D | 09:31 |
@sonney2k | sploving, so did it make a difference? | 09:33 |
@sonney2k | but try uninstalling openjdk first | 09:33 |
* sonney2k is catching the train | 09:34 | |
@sonney2k | read you in a few minutes | 09:34 |
sploving | sonney2k, sorry I use ubuntu, which does not have sun-java6-jdk | 09:36 |
sploving | how to see the path/classpath | 09:36 |
sploving | is set correct or not? | 09:37 |
blackburn | eh.. ubuntu has sun-java6-jdk | 09:37 |
sploving | oh | 09:40 |
sploving | i will have a try again | 09:40 |
blackburn | I have sun-java6-jdk on my ubuntu 10.10 | 09:43 |
sploving | me 10.04 | 09:43 |
blackburn | I think there is no difference, sun-java6-jdk should be available | 09:45 |
sploving | but I use "apt-get install sun-jave6-jdk", it said, it is replace by other similar software | 09:46 |
blackburn | just remove openjdk first | 09:46 |
sploving | i removed it | 09:46 |
blackburn | hm.. | 09:47 |
blackburn | what is the exact error when you are trying to install sun jdk? | 09:47 |
sploving | no erro, just no such package | 09:48 |
blackburn | hmm | 09:48 |
blackburn | what a strange thing | 09:49 |
sploving | I think openjdk is ok. maybe some path is not set correct | 09:50 |
blackburn | yeah I think so too | 09:51 |
blackburn | I didn't felt any difference with using both | 09:51 |
blackburn | just some EE issues but it is not the case | 09:51 |
blackburn | btw may be you could just set this paths when running with no use of global variables | 09:54 |
* sonney2k is back (train) | 10:01 | |
@sonney2k | sploving, I am compiling for openjdk now | 10:02 |
sploving | sonney2k, thx for your try | 10:02 |
@sonney2k | sploving, works without problems | 10:05 |
@sonney2k | just did | 10:05 |
@sonney2k | ./configure --interfaces=libshogun,java_modular | 10:05 |
@sonney2k | make | 10:05 |
@sonney2k | cd java_modular | 10:05 |
@sonney2k | javac *.java ../../examples/undocumented/java_modular/minimal_test.java | 10:05 |
@sonney2k | CLASSPATH=.:../../examples/undocumented/java_modular LD_LIBRARY_PATH=../libshogun:. java minimal_test | 10:05 |
@sonney2k | blackburn, if you have time could you try too? | 10:05 |
sploving | sonney2k, I will try | 10:06 |
blackburn | sonney2k: sorry I don't have now :( | 10:06 |
@sonney2k | blackburn, btw did you test the auPRC? does it return same outputs like the python script? | 10:06 |
blackburn | sonney2k: of course | 10:06 |
blackburn | 0.829476554611 | 10:07 |
blackburn | 0.829476554611 | 10:07 |
blackburn | copy-pasted from test output | 10:07 |
@sonney2k | blackburn, OK then I will just merge it :) | 10:07 |
blackburn | :) okay | 10:08 |
blackburn | last night have different auPRC but found the 'error' just took not proper trapezoid | 10:09 |
@sonney2k | blackburn, I think you should put the trapezoid thing into CMath later | 10:09 |
@sonney2k | I mean both auROC and auPRC need it... | 10:10 |
@sonney2k | something like compute_area() | 10:10 |
@sonney2k | sploving, so? | 10:10 |
blackburn | sonney2k: just noticed there is negative areas in PRC :) | 10:11 |
@sonney2k | blackburn, well done :D | 10:11 |
blackburn | just like on picture I attached, when precision is going 'left' | 10:11 |
blackburn | sonney2k: it is proper cause it will be just subtracted from auPRC and will return proper area | 10:12 |
CIA-90 | shogun: Soeren Sonnenburg master * r320094d / (5 files in 2 dirs): | 10:12 |
CIA-90 | shogun: Merged pull request #70 from SergeyLisitsyn/master. | 10:12 |
CIA-90 | shogun: PRCEvaluation class addition and some fixes for shogun.Evaluation - http://bit.ly/h3pnnC | 10:12 |
blackburn | not a bug but a feature :D | 10:12 |
blackburn | python script does it the same | 10:12 |
@sonney2k | negative area? impossible! | 10:12 |
blackburn | I could check it but now rushing for some work on game theory I should done :) | 10:14 |
@sonney2k | blackburn, yes l8r | 10:14 |
sploving | sonney2k, I will try another computer | 10:16 |
@sonney2k | sploving, please give the exact error if it does not work now | 10:17 |
@sonney2k | sploving, though that shouldn't be the problem... | 10:17 |
sploving | sonney2k, the error is: "Exception in thread "main" java.lang.UnsatisfiedLinkError: no Kernel in java.library.path" | 10:19 |
@sonney2k | which happens when? | 10:20 |
@sonney2k | sploving, ? | 10:20 |
sploving | at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1681) | 10:21 |
@sonney2k | ok could reproduce it: happens when I run CLASSPATH=../../examples/undocumented/java_modular:. LD_LIBRARY_PATH=../libshogun java minimal_test | 10:21 |
@sonney2k | instead of CLASSPATH=../../examples/undocumented/java_modular:. LD_LIBRARY_PATH=../libshogun:. java minimal_test | 10:21 |
sploving | yeap | 10:21 |
@sonney2k | which path are you in? | 10:21 |
@sonney2k | pwd ? | 10:21 |
sploving | /home/sploving/shogun/shogun/src/java_modular | 10:22 |
@sonney2k | it seems you didn't set LD_LIBRARY_PATH correctly | 10:22 |
@sonney2k | and there is a Kernel.so in there? | 10:22 |
sploving | yeap | 10:22 |
sploving | let me have a think | 10:24 |
@sonney2k | sploving, do you by chance overwrite LD_LIBRARY_PATH somewhere? | 10:24 |
@sonney2k | e.g. have it in your profile or .bashrc ? | 10:24 |
sploving | I edit the profile "classpath and path" | 10:25 |
@sonney2k | sploving, one way to rule that out would be to just copy libshogun*.so and Kernel.so to /usr/lib then try again | 10:25 |
@sonney2k | sploving, not path but ld_library_path | 10:25 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 240 seconds] | 10:30 | |
sploving | sonney2k, src/lua_modular/swig_typemaps.i: needs update | 10:31 |
sploving | fatal: Untracked working tree file 'src/lua_modular/shogun.lua' would be overwritten by merge. | 10:31 |
sploving | when I use another computer, and use "git pull" | 10:31 |
sploving | it said this | 10:31 |
blackburn | you have to checkout that file | 10:33 |
blackburn | (or to commit) :) | 10:34 |
sploving | most of time wast these meaning ful issue:( | 10:35 |
sploving | not meaningful | 10:35 |
sploving | :( | 10:35 |
sploving | sonney2k, we just put these issue aside. what is the next step? | 10:45 |
sploving | write a smll typemap | 10:45 |
sploving | ? | 10:45 |
@sonney2k | sploving, yes | 10:48 |
* blackburn has done damn game theory work | 10:48 | |
blackburn | sonney2k: will check logs when came home today | 10:48 |
@sonney2k | blackburn, ok | 10:49 |
blackburn | hope will not miss something important :) | 10:49 |
blackburn | see you | 10:49 |
-!- blackburn [~qdrgsm@188.168.5.38] has quit [Quit: Leaving.] | 10:50 | |
@sonney2k | sploving, I think a good start would be to write a typemap for e.g. Vector | 10:51 |
@sonney2k | like you did for lua already | 10:51 |
sploving | sonney2k, okay | 10:51 |
@sonney2k | so sploving get java to work - and then ping me back :-) | 10:54 |
@sonney2k | and in case you get stuck ask! | 10:57 |
@sonney2k | we have no time to lose especially this week :) | 10:57 |
sploving | sonney2k, when i make clean, i found that the generated files are not deleted | 10:57 |
sploving | and .o | 10:57 |
@sonney2k | ? | 10:57 |
@sonney2k | sploving, what do you want to do? | 10:58 |
@sonney2k | erase all files that are not in your git repo? Then use git clean -dfx | 10:58 |
sploving | yep | 10:59 |
@sonney2k | did that solve your problem? | 11:00 |
@sonney2k | please always give feedback - otherwise I don't know what you are doing / where you are stuck... | 11:00 |
sploving | okay. | 11:01 |
@sonney2k | so what are you doing now? | 11:03 |
sploving | I tried to cp all the files to another computer | 11:03 |
sploving | as when I use the git clone, it is very slow | 11:04 |
@sonney2k | sploving, from github? Shogun's repository is only 20 MB... | 11:04 |
sploving | 1kb/s | 11:05 |
@sonney2k | that is slow... | 11:06 |
* sonney2k wonders why | 11:06 | |
-!- sploving [~root@124.16.139.196] has quit [Ping timeout: 240 seconds] | 11:10 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 11:12 | |
sploving | Receiving objects: 0% (257/55876), 51.99 KiB | 3 KiB/s | 11:12 |
sploving | Receiving objects: 0% (341/55876), 67.99 KiB | 1 KiB/s | 11:12 |
-!- f-x [~gen@180.149.49.227] has joined #shogun | 12:05 | |
-!- f-x [~gen@180.149.49.227] has quit [Ping timeout: 252 seconds] | 12:20 | |
-!- f-x [~gen@180.149.49.227] has joined #shogun | 12:22 | |
-!- f-x [~gen@180.149.49.227] has quit [Client Quit] | 12:27 | |
-!- f-x [~user@180.149.49.227] has joined #shogun | 12:28 | |
-!- f-x` [~user@180.149.49.227] has joined #shogun | 12:38 | |
-!- heiko [~heiko@dhcp-ce170189.zmb.uni-duisburg-essen.de] has joined #shogun | 12:39 | |
heiko | hi there | 12:39 |
CIA-90 | shogun: Soeren Sonnenburg master * r7ce4b0e / (7 files in 3 dirs): Add support for block norm mkl - http://bit.ly/dOKYFP | 12:41 |
CIA-90 | shogun: Soeren Sonnenburg master * r8af2ece / (2 files): fix naming clash renaming degree -> m_degree - http://bit.ly/fpmLJ5 | 12:41 |
CIA-90 | shogun: Soeren Sonnenburg master * r05b88ec / (7 files in 4 dirs): | 12:41 |
CIA-90 | shogun: add an example for mixed norm mkl and make BLOCK_NORM naming more | 12:41 |
CIA-90 | shogun: consistent (always with '_') - http://bit.ly/if8PW5 | 12:41 |
-!- f-x` [~user@180.149.49.227] has quit [Client Quit] | 12:41 | |
-!- f-x [~user@180.149.49.227] has quit [Ping timeout: 240 seconds] | 12:42 | |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 12:43 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has joined #shogun | 12:43 | |
Tanmoy | hi all | 12:44 |
heiko | hi | 12:44 |
f-x | hey! | 12:44 |
-!- sploving1 [~sploving@124.16.139.198] has joined #shogun | 13:12 | |
heiko | sonney2k, are you there? | 13:17 |
-!- sploving [~root@124.16.139.196] has left #shogun [] | 13:38 | |
-!- sploving1 [~sploving@124.16.139.198] has quit [Ping timeout: 258 seconds] | 13:42 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 13:43 | |
@sonney2k | heiko, yes now :) | 13:59 |
heiko | sonney2k, ok :) | 13:59 |
heiko | You pointed me to the read/write function of the Parameter class, which I checked and understood. | 13:59 |
heiko | However, Iam still not really sure, how you want the setting/getting of parameters to look like. | 13:59 |
heiko | Currently (in the poc code), every parameter is accessible via | 13:59 |
heiko | TParameter* Parameter::get_parameter(int32_t idx) | 13:59 |
heiko | so in principle, it is possible to manipulate the void* m_parameter member of TParameter. | 13:59 |
heiko | Members other than the m_parameter do not need to be changed (name, description, type etc) | 13:59 |
heiko | I think, currently, one could do something like | 14:00 |
heiko | *(gaussian_kernel->get_cross_val_parameter(some_index)->m_parameter) = new_value | 14:00 |
heiko | However, there is no type checking this way. Thats the point where I am unsure. | 14:00 |
-!- heiko was kicked from #shogun by bettyboo [flood] | 14:00 | |
-!- heiko [~heiko@dhcp-ce170189.zmb.uni-duisburg-essen.de] has joined #shogun | 14:00 | |
heiko | Do you want a method like | 14:00 |
heiko | set_cross_validation_parameter(int32_t idx, void* m_parameter) | 14:00 |
heiko | which handles all the type stuff as done in the read/write methods of Parameter? Or how should this work? | 14:00 |
@mlsec | bettyboo: did you just kick heiko for no reason?! | 14:01 |
@bettyboo | mlsec: just not commit next thing before pull will be closed :) | 14:01 |
@mlsec | bettyboo: i thought he commited? | 14:02 |
@bettyboo | mlsec: I have implemented a fast nearest neighbor classification (in case of k=1) in CKNN class. what should I do after cleaning and commenting to get it commited? | 14:02 |
@bettyboo | zzz mlsec | 14:02 |
@mlsec | hmm. that's not too funny. | 14:02 |
@mlsec | :) | 14:02 |
heiko | (sonney2k, I just wrote this into an email :) | 14:03 |
@bettyboo | heiko, yeah | 14:03 |
heiko | (...but now you are here) | 14:03 |
@sonney2k | heiko, I think we should do this name based | 14:04 |
heiko | think so | 14:05 |
@sonney2k | or at least have a function that maps name -> idx | 14:05 |
heiko | ok, but how should this work? | 14:05 |
heiko | set_param(char* name, void* content) | 14:05 |
heiko | ? | 14:05 |
@sonney2k | yes but not void* but the real type | 14:07 |
heiko | so we need a setter for every possible type? | 14:07 |
heiko | and the function checks if this type matches the type of the parameter? | 14:07 |
-!- f-x_ [b49531e3@gateway/web/freenode/ip.180.149.49.227] has joined #shogun | 14:09 | |
heiko | if you have parameter "A" with type X and call set_parameter("A", something_of_type_y) .... | 14:11 |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 252 seconds] | 14:13 | |
-!- f-x_ is now known as f-x | 14:14 | |
@sonney2k | heiko, I tend to say yes - I am not sure yet - at some point we have to check types | 14:14 |
heiko | but one needs to know the type of the parameter one wants to set | 14:15 |
@sonney2k | heiko, we should first start with a set_parameter function for a scalar | 14:15 |
heiko | ok | 14:15 |
@sonney2k | heiko, yes and so a check makes sense | 14:15 |
@sonney2k | scalar float64_t that is | 14:15 |
@sonney2k | then I guess CSGObject* is the next level | 14:16 |
heiko | this should be quite easy, far easier than these read/write routines | 14:16 |
heiko | so should this method be added to TParameter? or to Parameter? | 14:18 |
@sonney2k | TParameter | 14:19 |
@sonney2k | heiko, I am thinking of whether one should do it totally differently | 14:20 |
heiko | that is? | 14:20 |
@sonney2k | I mean say the modsel object constructs a new TParameter object | 14:20 |
@sonney2k | with datatype ptr to parameter name etc | 14:21 |
@sonney2k | and then in SGObject we add a function that sets the parameters from TParameter | 14:21 |
@sonney2k | a parameter from a TParameter | 14:22 |
@sonney2k | that is | 14:22 |
heiko | correct me if I am wrong: doesnt TParameter::m_parameter is a pointer to the member variable of a class? | 14:24 |
heiko | if you then replace a TParameter by a new one, has something changed? | 14:25 |
heiko | if this pointer is euqal? | 14:25 |
heiko | I mean if parameters are set, only the value changes, the rest stays the same, so why replace all? | 14:25 |
@sonney2k | We have m_parameters in CSGObject for all parameters | 14:26 |
@sonney2k | Now one could add a function to CSGObject that gets a single TParameter object (not to confuse with the class Parameter) and we use this to tell the Parameter object m_parameters to set the value of *one* paramter | 14:27 |
@sonney2k | so I don't want to replace pointers but use the TParameter interface to set parameters from other TParameters | 14:31 |
@sonney2k | heiko, does that make any sense? | 14:31 |
heiko | I am unsure if I understood what you mean | 14:32 |
heiko | this function is already there (to get a TParameter object from CSGObject) | 14:33 |
heiko | how would a call look like? | 14:34 |
heiko | i mean of the whole procedure | 14:34 |
@sonney2k | how would one get a TParameter object form CSGObject? | 14:34 |
@sonney2k | via m_parameters you mean? | 14:34 |
heiko | no thats clear, I meant the rest | 14:35 |
@sonney2k | anyway, it would work similar to how loading saving is done now: | 14:35 |
@sonney2k | currently one calls save() and the object is dumped in whatever format to e.g. disk | 14:36 |
@sonney2k | now one could create a second parameter object pointing at the actual parameter to be used and then 'save' it and load it again | 14:37 |
heiko | yes | 14:38 |
sploving | where is the irc log url? | 14:40 |
@sonney2k | heiko, if you don't feel comfortable with this we can as well do it the 'simple' way - but we still need a way to tell that we have parameters of certain type anyways | 14:40 |
@sonney2k | sploving, #topic :-) | 14:41 |
sploving | #topic | 14:41 |
heiko | but when you save or load, you do not know anything about the parameter beforehand, while in the context of setting cross-val-parameters, only the actual value has to be changed. | 14:41 |
sploving | i see | 14:41 |
@sonney2k | heiko, I don't see the difference | 14:42 |
@sonney2k | when you load, you already have an object there | 14:42 |
@sonney2k | so you only change the value of the variable | 14:42 |
heiko | yes, true, sorry | 14:43 |
@sonney2k | If you don't feel comfortable with that approach I am also fine with doing it the more direct way that is setting an actual parameter value by name | 14:43 |
@sonney2k | and value of course :) | 14:43 |
heiko | its not that I feel unconfortable with it, but I havent understood it completely :) | 14:44 |
@sonney2k | heiko, ok the idea would be that the model selection object creates via whatever means another CParameter object pointing to all the actual model selection parameters | 14:46 |
heiko | yes clear. | 14:47 |
@sonney2k | then one could set say m_model_selction_parameters->set_from_parameter(params) | 14:47 |
@sonney2k | and then there is some loop inside going through all of the parameters setting all the values | 14:47 |
@sonney2k | and checking types | 14:47 |
@sonney2k | so it is in fact very similar to what load/save do | 14:48 |
heiko | and the argument of set_from_parameter(params) is ? | 14:49 |
@sonney2k | params? A CParameter instance. | 14:49 |
heiko | and it contains all parameters that one wants to set? the values of all elements of this params instance are copied to the corresponding elements of m_model_selection_parameters | 14:53 |
@sonney2k | params instance has just pointers to the actual variables with their values, so the set_from* function will set the values to the ones it has | 14:54 |
heiko | alright | 14:54 |
heiko | understood it :) | 14:54 |
heiko | and the model selection object creates this instance ... | 14:55 |
heiko | and therefore it needs all the type information, and therefore there has to be a getter for the TParameter elements of CParameter | 14:55 |
@sonney2k | yes and it must somehow know the parameters anyways | 14:56 |
heiko | true | 14:56 |
heiko | ok, thanks for your patience ;) | 14:56 |
-!- f-x [b49531e3@gateway/web/freenode/ip.180.149.49.227] has quit [Ping timeout: 252 seconds] | 14:56 | |
@sonney2k | the alternative is of course to do all the setters / getters as extra functions - also valid | 14:57 |
heiko | I think the second method is quite elegant | 14:57 |
-!- vojta [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has joined #shogun | 14:57 | |
-!- vojta is now known as VojtechFranc | 14:57 | |
@sonney2k | but this approach here would just need pointers so we avoid copying | 14:57 |
* sonney2k checks his watch | 14:57 | |
@sonney2k | we will be having the kick-off meeting soon | 14:57 |
-!- mikiobraun [~mikio@ally.ml.tu-berlin.de] has joined #shogun | 14:58 | |
heiko | ah, sorry, forgot to tell this, I cannot be there. Family business. I have to go in a few minutes | 14:58 |
@sonney2k | heiko, ok - read the chatlog then | 14:59 |
heiko | will do | 14:59 |
heiko | See you tomorrow! | 14:59 |
@sonney2k | heiko, OK heiko cu! | 15:00 |
-!- JohnL_ [6c15273f@gateway/web/freenode/ip.108.21.39.63] has joined #shogun | 15:01 | |
-!- heiko [~heiko@dhcp-ce170189.zmb.uni-duisburg-essen.de] has left #shogun [] | 15:01 | |
-!- f-x [~gen@210.212.58.229] has joined #shogun | 15:02 | |
@sonney2k | Lets start in 1 minute. | 15:03 |
@sonney2k | Alright. Hello all! | 15:04 |
@sonney2k | Welcome to this first rather spontaneous kick off meeting. I had to schedule it that early as I will be away most likely from Tuesday next week (expecting a new baby...) and I would like that everyone gets to know his student/mentor before I am away/less responsive. | 15:05 |
@sonney2k | I think except for sploving we all participate in GSoC for the first time. So lets make sure everyone read the advice for students http://code.google.com/p/google-summer-of-code/wiki/AdviceforStudents and http://www.booki.cc/gsoc-mentoring/_v/1.0/best-practices/ for mentors respectively. | 15:05 |
-!- mikiobraun [~mikio@ally.ml.tu-berlin.de] has quit [Ping timeout: 246 seconds] | 15:05 | |
@sonney2k | Since we are all new it is especially important that we finish with 100% success. So it is better to start early and to regularly stay in contact with your students/mentors. | 15:06 |
JohnL_ | I read a portion of it---actually pretty good. | 15:06 |
@sonney2k | I would propose to once you have started (mentors <-> students please discuss when it is a good time!) to in the beginning at least once per day have a short chat (IRC, ...). Later (say after 2 weeks of more intense collaboration) a meeting once per week is most likely sufficient. | 15:06 |
@mlsec | hiho | 15:07 |
@sonney2k | Is that feasible? | 15:07 |
@sonney2k | Apart from that, please use whenever possible public channels, e.g. communicate via the mailinglist or IRC here and CC your mentor/student. | 15:08 |
@sonney2k | Then for the workflow - I suggest that you all get yourself an account at github and fork shogun. | 15:08 |
@sonney2k | You then plan and discuss how to proceed. | 15:08 |
sploving | mlsec, are you my mentor? | 15:09 |
JohnL_ | Shashwat is working with both you and me. If we can email back and forth at least once a day, I think we can start well. | 15:09 |
@mlsec | sploving: no, chris will take care as primary mentor | 15:09 |
@sonney2k | sploving, mikiobrain is yours | 15:09 |
f-x | I agree, JohnL_ and sonney2k | 15:09 |
@mlsec | oh, sorry. | 15:09 |
@mlsec | ;) | 15:09 |
@bettyboo | strange | 15:09 |
@sonney2k | JohnL_, f-x I agree... I am hoping that f-x has his first patch in shogun this week :) | 15:10 |
@sonney2k | Anyway, then whenever a student has a simple patch ready he issues a pull request and you will be notified. | 15:10 |
@sonney2k | You use github to comment on his patch and potentially meet in IRC etc discuss more difficult issues. Once everything looks nice and shiny - tell me to do the merge. | 15:10 |
@sonney2k | Your probably all know the timeline http://www.google-melange.com/gsoc/events/google/gsoc2011 | 15:12 |
@sonney2k | Mid-term evaluatins deadline is July 15 - so while that looks like ample time it is simply better to getting things done early. | 15:13 |
@sonney2k | To avoid the deadline hectic close to the deadline... | 15:13 |
@sonney2k | Does anyone have any questions regarding to that? | 15:13 |
sploving | if we finished the project before mid_term evaluation, then we could have holidays after that? | 15:14 |
JohnL_ | none here | 15:14 |
@sonney2k | sploving, holidays are always possible - of course not for the full period :) | 15:15 |
VojtechFranc | sonney2k, provided I want to communicate with the student by email is it reasonable to spam others using mailinglist? if yes, which one? | 15:15 |
@sonney2k | and you should early on notify us when you are planning to do those - I guess best time period is directly after the mid term evaluation deadline | 15:15 |
@sonney2k | VojtechFranc, please spam the mailinglist yes | 15:16 |
@sonney2k | I think it is best that every student / mentor at least idles in this IRC channel - live discussions can be done here too and there are channel logs. I find this a very convenient way to make progress. | 15:18 |
sploving | I did not see my mentor | 15:18 |
@sonney2k | Of course sometimes you may want to discuss things in private - put public channels are preferred | 15:18 |
VojtechFranc | sonney2k, sorry for stupid question, which mailinglist should we use ? | 15:19 |
JohnL_ | I general prefer email to IRC, as too much of my day is not interruptible. | 15:19 |
-!- mikiobraun [~mikio@squid.ml.tu-berlin.de] has joined #shogun | 15:20 | |
@sonney2k | JohnL_, it is up to the mentor... | 15:20 |
@sonney2k | sploving, there mikiobraun is | 15:21 |
@sonney2k | your mentor | 15:21 |
sploving | hello mikiobraun | 15:21 |
mikiobraun | sorry, my internet connection dropped without telling me :( | 15:21 |
@mlsec | Besides IRC, the public mailing list is also an option for generic discussion and coordination | 15:21 |
@sonney2k | That is all I wanted to say - now please ask plenty of questions. | 15:22 |
@mlsec | how is the progress of each student documented? via github commits? | 15:22 |
-!- Ziyuan [~Ziyuan@116.21.255.174] has joined #shogun | 15:23 | |
@sonney2k | In the end patches are the only way to measure whether a student is active or not. | 15:23 |
@sonney2k | So I really recommend to issue pull requests early and often (that don't break things). | 15:24 |
@mlsec | I would also count activity in the branch of a student, as he might prefer to work for a longer time on its own code before sending out a patch | 15:24 |
@sonney2k | Please don't deliver a finished product at the end of GSoC but many small patches towards achieving your goals. | 15:24 |
sploving | I would like finish one features then submit the patch | 15:25 |
serialhex | morning all... err afternoon to everyone :P | 15:25 |
@sonney2k | sploving, well if you for example finish a typemap for e.g. vectors - you should already submit that | 15:26 |
-!- Guest [53173991@gateway/web/freenode/ip.83.23.57.145] has joined #shogun | 15:26 | |
@sonney2k | not only when you did all the typemaps for java and ported examples for example | 15:26 |
@sonney2k | Any further questions? | 15:27 |
sploving | that is right | 15:27 |
@mlsec | are there any fixed or periodic meetings during the next months? | 15:28 |
@sonney2k | I think we should have a short meeting with everyone involved once per month | 15:30 |
@sonney2k | I would suggest to have the next meeting on May 27 | 15:30 |
@sonney2k | ahh and I forgot one important thing | 15:31 |
@sonney2k | progress reports: I would very much welcome that every student would write a weekly *short* email about his progress. | 15:31 |
@sonney2k | starting from the official beginning of GSoC only that is. | 15:31 |
serialhex | i also believe some people were talking about blogging about their project too... | 15:32 |
@mlsec | why not? | 15:32 |
@sonney2k | everyone who has a blog / wants to blog about his progress (in more detail) is very welcome to. I've set up a planet for that so just tell me your rss feed and you will be listed at http://shogun-toolbox.org/planet/ | 15:33 |
@sonney2k | If there are no further questions I would then close this meeting. | 15:34 |
f-x | will the blog be used as a way to gauge progress? | 15:35 |
@sonney2k | ...and thank everyone for attending! Mark May 27 13:00 UTC in your calendars ... and meet you soon again :) | 15:35 |
@sonney2k | f-x, we will use that too but ultimately it is accepted patches that count | 15:35 |
f-x | sonney2k: that's fine.. | 15:36 |
JohnL_ | ==Soeren | 15:36 |
sploving | hello sonney2k, I asked my mentor about that the problem. and solved it ! | 15:36 |
sploving | I need rename Kernel.so to libKernel.so | 15:37 |
@sonney2k | JohnL_, yes? | 15:37 |
sploving | It works! | 15:37 |
@sonney2k | sploving, ahh yes sorry :) | 15:37 |
@sonney2k | great :) | 15:37 |
JohnL_ | Just expressing agreement. | 15:37 |
@sonney2k | alright then case dismissed! | 15:37 |
@mlsec | i am fine | 15:38 |
Guest | Hi everyone. I hope I am not interrupting in your discussion ;) | 15:38 |
Guest | Could anyone please tell me if I can define in shogun my own distance for kmeans? | 15:38 |
-!- JohnL_ [6c15273f@gateway/web/freenode/ip.108.21.39.63] has quit [Quit: Page closed] | 15:39 | |
@mlsec | ;> | 15:39 |
@sonney2k | Guest, sure, http://www.shogun-toolbox.org/doc/classshogun_1_1CDistance.html | 15:39 |
@sonney2k | just write your own in C++ | 15:40 |
@sonney2k | there are enough exampels in the docs for it | 15:40 |
f-x | sonney2k: why will 2 fread()s be required for label and data? | 15:42 |
@sonney2k | f-x, great to see you back and that your internet connection is more stable today :) | 15:43 |
f-x | presently, i read till \n using getline | 15:43 |
f-x | yeah :) got a headache last night!! | 15:43 |
@sonney2k | sploving, great - so now just get some vector typemap to work for a starter and all good | 15:43 |
@sonney2k | f-x, ahh for ascii files you won't need that | 15:43 |
@sonney2k | for binary files you might have the label to be a double and your data to be bytes | 15:44 |
f-x | right. but i guess i'll make the ascii version proper first.. | 15:44 |
@sonney2k | f-x, yes yes! | 15:44 |
f-x | what other changes do i need to make now? | 15:45 |
f-x | to get my first patch :P | 15:45 |
f-x | my understanding is that the main concern is the single block of memory for the buffer | 15:45 |
@sonney2k | f-x, push and pull request again - I will then have a more close look - could be that I want more changes later even though I am merging it now | 15:45 |
@sonney2k | f-x, no the copying | 15:46 |
f-x | the copying i removed in my last commit | 15:46 |
sploving | sonney2k, mikiobraun suggest me use ant build the basic environment. then I will deal with typemap | 15:46 |
serialhex | sploving: when do you think you are going to start on the ruby typemaps? | 15:46 |
@sonney2k | serialhex, I think that was last in his schedule | 15:46 |
f-x | and made it necessary to call free_feature_vector from outside | 15:46 |
@sonney2k | f-x, yes :) | 15:46 |
serialhex | hmm... ok | 15:47 |
sploving | I think I should finish java first | 15:47 |
@sonney2k | serialhex, no one minds you starting with this - sploving could also spent his time on csharp later :) | 15:47 |
f-x | sonney2k: i too think there will be many changes to be made later.. i've hardly made it as efficient as i would like | 15:47 |
serialhex | ahh... so he beat me by 3 languages instead of just 2!!! :P | 15:48 |
serialhex | sploving: would you mind if i did that? started on the ruby typemaps? | 15:48 |
sploving | serialhex, your are also interested in the binding? | 15:48 |
@sonney2k | serialhex, no - but if you want to do ruby he could do csharp for example | 15:49 |
@sonney2k | mikiobraun, why don't you talk with sploving in the public? | 15:49 |
serialhex | yeah, my project was the ruby typemaps sploving | 15:49 |
f-x | sploving is going to get his early holiday, it looks like :) | 15:49 |
@bettyboo | funny | 15:49 |
serialhex | and she gets smarter and smarter :P | 15:49 |
sploving | sonney2k, mikiobraun have another meeting now | 15:49 |
@sonney2k | sploving, I will check if I can prepend the lib prefix to shogun's java .so objects | 15:50 |
sploving | serialhex, it does not matter. If you are interested, please do it | 15:50 |
serialhex | ok, thanks sploving!!! | 15:50 |
sploving | thx. sonney2k | 15:50 |
sploving | serialhex, you are a ruby player? | 15:51 |
sploving | I thouth you maybe are very familar with it | 15:51 |
mikiobraun | sonney2k: we just catched up on what I've missed | 15:52 |
serialhex | sploving, it's the only language i'm really good at | 15:52 |
serialhex | i'm not an expert by any measure, but i can program my way out of a box sploving :D | 15:52 |
sploving | then are you familiar with swig? | 15:53 |
Guest | @sonney2k: Thanks for the tips | 15:56 |
sploving | serialhex, keep in touch! | 15:56 |
serialhex | sploving: i've been studying it since i found out about this project, and have done a fw things with it | 15:56 |
serialhex | will do sploving, are you on github? | 15:57 |
sploving | ok. then we could learn from each other | 15:57 |
serialhex | excellent, what time zone are you in?? | 15:57 |
sploving | yep. I submitted a small lua typemap | 15:57 |
sploving | UTC+8 | 15:58 |
sploving | what about you | 15:58 |
serialhex | -5, new york is in my timezone | 15:58 |
* serialhex cant stand this wonky clock with only cities and no actual *timezones*!!!! | 15:59 | |
@sonney2k | sploving, serialhex would be cool if you interact ... I guess serialhex could do a few typemaps for ruby... | 15:59 |
sploving | sonney2k, i see. | 15:59 |
@sonney2k | serialhex, btw it is all prepared --- you only need ./configure --interfaces=libshogun,ruby_modular | 15:59 |
sploving | serialhex, you could just write a small vector typemap | 16:00 |
sploving | for a begin | 16:00 |
serialhex | sploving: could you give me a major city in your country because i can only select from regions/cities | 16:01 |
sploving | China | 16:01 |
sploving | beijing | 16:01 |
Guest | sonney2k: I took a look on that website you sent. Unfortunately, it is still not clear for me which distance should I inherit? CSimpleDistance, CRealDistance, CSparseDistance or CCustomDistance? Could you please give me any hint? | 16:02 |
serialhex | sploving: ok, i'll see what i can come up with. my C++ isnt very good, though i have submitted a few patches (gaussian kernel & normal distributed random numbers) but i'll check with you before doing any pull requests | 16:02 |
Guest | sonney2k: I would like to implement a distance (using vectors of doubles) for kmeans. | 16:02 |
@sonney2k | what types do your features have | 16:02 |
@sonney2k | then derive from CSimpleDistance | 16:02 |
Guest | sonney2k: thanks | 16:03 |
serialhex | sploving: it's ~10 pm there? | 16:03 |
sploving | serialhex, yeap | 16:03 |
serialhex | you're 12 hours ahread of me then :D | 16:03 |
@sonney2k | I guess serialhex can start now and sploving can take over - 100% efficiency for a 24hr day :) | 16:04 |
serialhex | lol :P | 16:04 |
Guest | sonney2k: Just one more question if I can. Is there any distance defined between 2 covariance matrices? | 16:05 |
@sonney2k | Guest, not that I know of | 16:05 |
Guest | sonney2k: ok, thanks a lot for info | 16:06 |
serialhex | sploving: what are you known as on github? | 16:06 |
sploving | sploving | 16:06 |
sploving | u? | 16:06 |
serialhex | serialhex - i was looking for sploving but couldnt find it | 16:07 |
-!- Guest [53173991@gateway/web/freenode/ip.83.23.57.145] has left #shogun [] | 16:07 | |
* serialhex needs better glasses | 16:07 | |
sploving | I am one of the people who "watches" shogun | 16:09 |
serialhex | yeah i found your fork | 16:09 |
serialhex | & i'm watching you | 16:09 |
@sonney2k | f-x, https://gist.github.com/511ebb332a6f18d02c62/1d97195bb0d1ca258eae680e03f8fc479de458f3 looks nice! | 16:11 |
f-x | sonney2k: :) hopefully i can make that CStreamingFeatures constructor a bit tidier | 16:12 |
-!- mikiobraun [~mikio@squid.ml.tu-berlin.de] has left #shogun [] | 16:12 | |
sploving | serialhex, when are you going to start ruby binding? | 16:12 |
serialhex | later on today perhaps sploving, i'll make a new branch and you can check my progress there. i'll also send you messages via github or through e-mail if you like | 16:13 |
sploving | github is better. everyone could see it | 16:14 |
serialhex | ok | 16:15 |
sploving | could check it | 16:15 |
serialhex | i'm very new to *actual* programming, i've studied a lot of the theory before, so you might have to bear with me a bit on this... i'm good at running headlong into brick walls :D | 16:15 |
serialhex | and then it sucks when people tell me that the door was 1m to my right :P | 16:16 |
sploving | It does not matter, just try your best. then you could be an expert programmer | 16:18 |
serialhex | thats what i'm working on! | 16:18 |
sploving | I have to go now. Looking forward to hearing from your progress! | 16:19 |
serialhex | ok, ttyl sploving! | 16:21 |
sploving | bye | 16:21 |
@bettyboo | over and out | 16:21 |
-!- VojtechFranc [~quassel@2001:718:2:1634:221:6aff:fe58:151a] has quit [Remote host closed the connection] | 16:22 | |
@sonney2k | f-x, seen JL's comment - as I told :D | 16:22 |
@sonney2k | sploving, I just did the prefixing with lib | 16:23 |
CIA-90 | shogun: Soeren Sonnenburg master * r4763b14 / src/libshogun/classifier/AveragedPerceptron.cpp : fix a variable naming clash - http://bit.ly/hCXLwD | 16:24 |
CIA-90 | shogun: Soeren Sonnenburg master * rc8e0065 / src/configure : prefix java_modular generated .so files with lib - http://bit.ly/eAC0Sv | 16:24 |
serialhex | sonney2k: who made bettyboo?? mlsec???? is he the one to blame for humanity's downfall to the robots??? | 16:25 |
@bettyboo | serialhex: so this worked like this: every SGObject class may register parameters to an extra list for cross-val | 16:25 |
serialhex | she's hiding her intellect now :D | 16:25 |
@sonney2k | bettyboo, don't remind me of that! | 16:25 |
@bettyboo | serialhex, lolomat | 16:25 |
@bettyboo | sonney2k: hm! good you remind me | 16:25 |
f-x | sonney2k: hmm.. that is a point i would definitely like to discuss with both of you.. get that buffer thing straightened out once and for all! :) | 16:25 |
serialhex | yeah... we're doomed | 16:26 |
@sonney2k | f-x, the question is how would you want to proceed. I guess first get a minimal patch into shogun - then polish this until it is lightning fast and then work on SGD (and related online method integration) right? | 16:27 |
f-x | sonney2k: that was exactly the plan.. | 16:27 |
@sonney2k | I mean you could do binary features too sockets etc etc - so spend a month on streaming features | 16:27 |
@sonney2k | serialhex, mlsec only - he loves such stuff - and probably just got texted - hah :D | 16:28 |
f-x | but i want to get it done properly for a wider range of input first - especially sparse features | 16:28 |
f-x | for that i need clarification on a few points, and then i could get a basic version supporting sparse features pretty quickly | 16:28 |
@sonney2k | yeah makes sense | 16:28 |
@sonney2k | serialhex, sploving is gone by now right? | 16:29 |
CIA-90 | shogun: Soeren Sonnenburg master * raad532b / (2 files): fix (member) variable naming clashes - http://bit.ly/f42pQh | 16:29 |
serialhex | yeah sonney2k, i want to know how he did it... that's my main goal in learning all of this - A.I. | 16:29 |
CIA-90 | shogun: Soeren Sonnenburg master * re1dcd70 / (2 files): fix member naming clash - http://bit.ly/hmUNep | 16:29 |
serialhex | sonney2k: i believe so yes | 16:29 |
@sonney2k | serialhex, just simple rules imho ... | 16:30 |
sploving | sonney2k, not go yet | 16:30 |
sploving | I meant to go and saw your patch | 16:30 |
@sonney2k | sploving, ah - just wanted to tell you that I am now doing the lib prefixing | 16:30 |
sploving | i am trying | 16:31 |
@sonney2k | I totally forgot about that... | 16:31 |
sploving | sonney2k, it does not generate the lib | 16:35 |
sploving | i just use make clean, make again | 16:35 |
sploving | i know | 16:35 |
sploving | i should configure | 16:36 |
@sonney2k | sploving, you did git pull | 16:38 |
@sonney2k | ? | 16:38 |
-!- sploving [~root@124.16.139.196] has quit [Read error: Operation timed out] | 16:42 | |
-!- sploving [~root@124.16.139.196] has joined #shogun | 16:42 | |
@sonney2k | $ ll java_modular/*.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 1.3M Apr 27 16:45 java_modular/libClassifier.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 475K Apr 27 16:45 java_modular/libClustering.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 603K Apr 27 16:45 java_modular/libDistance.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 683K Apr 27 16:45 java_modular/libDistribution.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 567K Apr 27 16:45 java_modular/libEvaluation.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 9.7M Apr 27 16:46 java_modular/libFeatures.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 2.5M Apr 27 16:45 java_modular/libKernel.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 2.7M Apr 27 16:45 java_modular/libLibrary.so | 16:48 |
@sonney2k | -rwx------ 1 sonne sonne 789K Apr 27 16:45 java_modular/libPreProc.so | 16:48 |
-!- sonney2k was kicked from #shogun by bettyboo [flood] | 16:48 | |
-!- sonney2k [~shogun@87.118.92.43] has joined #shogun | 16:48 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 16:48 | |
@sonney2k | bettyboo, grrh%^@!!!!!!!!!#$@@@@%^ | 16:49 |
@bettyboo | sonney2k: well its actually mentioned that there are a whole lot of preprocessing methods available that would complement the features in shogun | 16:49 |
serialhex | lol... maybe you can make it so bettyboo can't boot mods?? | 16:50 |
@bettyboo | serialhex: I have to reboot the machines(now i am on windows) | 16:50 |
serialhex | i dont think you're on windows bettyboo... just call it a hunch | 16:50 |
@bettyboo | serialhex: eh! it's cold on street | 16:51 |
sploving | sonney2k, we should also definitely put all the Java classes into a package "shogun". Kernel in package shogun needs to be defined in shogun/Kernel.java | 16:53 |
sploving | swig has a opt to define the package | 16:53 |
@sonney2k | sploving, yeah but that is rather easy with swig | 16:53 |
@sonney2k | like doing an ant packe etc | 16:53 |
sploving | yeap. what about the classpath and Ld_library_path | 16:54 |
@sonney2k | what is still unknown for now is how to do a typemap - if you know how to do that with java - everything is set - so if you look at java's vector class and can wrap that then all good | 16:54 |
@sonney2k | sploving, there is no proper make install yet | 16:54 |
@sonney2k | sploving, look at Makefile.template line 286 | 16:55 |
@sonney2k | nothing there yet... | 16:55 |
sploving | okay. I know. then I will do a small typemap this week | 16:55 |
@sonney2k | sploving, thanks! | 16:58 |
sploving | not at all | 16:59 |
sploving | sonney2k, see you | 17:00 |
@sonney2k | sploving, ok l8r | 17:01 |
sploving | l8r | 17:01 |
f-x | sonney2k: is the new operator is overloaded for shogun objects? | 17:02 |
f-x | (extra 'is' above) | 17:02 |
serialhex | sonney2k: so is there anything additional you would like to say to me? | 17:02 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has joined #shogun | 17:21 | |
-!- blackburn [~qdrgsm@188.168.2.98] has joined #shogun | 17:26 | |
@sonney2k | blackburn, enough game theory for the day? | 17:28 |
blackburn | sonney2k: yeah, passed small mid-term 'test' | 17:28 |
blackburn | now reading logs | 17:28 |
@sonney2k | blackburn, I've had a look at your PRC figure - bad indeed | 17:28 |
blackburn | bad what? | 17:29 |
blackburn | it is similar to python one.. | 17:29 |
@sonney2k | I mean negative area is not correct | 17:29 |
@sonney2k | not sure how you even constructed that | 17:29 |
@sonney2k | blackburn, yes I know - even worse! | 17:30 |
@sonney2k | could mean that a lot of stuff we were doing is just crap | 17:30 |
blackburn | sonney2k: eh.. I'm disappointed with it | 17:30 |
blackburn | what should I do? | 17:30 |
@sonney2k | blackburn, that is the problem - I don't know we used this ourselves | 17:31 |
blackburn | sonney2k: I see.. | 17:32 |
blackburn | sonney2k: read logs. seems like my mentor wasn't there | 17:32 |
@sonney2k | can you avoid negative areas when you compute things from the y axis | 17:32 |
blackburn | sonney2k: hmm but | 17:33 |
blackburn | a minute | 17:33 |
@sonney2k | blackburn, cwidmer? | 17:33 |
@sonney2k | No idea | 17:33 |
@sonney2k | he said he wanted to come. | 17:33 |
blackburn | sonney2k: but there is no neg areas! | 17:33 |
blackburn | or.. | 17:34 |
blackburn | oh I'm tired.. have to eat something :) | 17:34 |
blackburn | sonney2k: so that picture I attached is bad one? | 17:34 |
@sonney2k | no there are not | 17:34 |
@sonney2k | enjoy your meal :) | 17:34 |
blackburn | sonney2k: but what is the problem? | 17:35 |
@sonney2k | blackburn, the picture - I don't know - could be correct even | 17:35 |
blackburn | http://dl.dropbox.com/u/10139213/share/prc_test.png | 17:35 |
blackburn | here it is | 17:35 |
@sonney2k | the negative area | 17:35 |
blackburn | there is no neg area cause trapeziod is 'rotated' | 17:35 |
@sonney2k | I thought you said you had a negative area? | 17:36 |
blackburn | sonney2k: I was wrong :) | 17:36 |
blackburn | sorry :) | 17:36 |
@sonney2k | if you rotate it I would have hope that it is ok - because higher recall can only mean lower precision | 17:37 |
@sonney2k | so the precision in this figure is 60-70 percent? | 17:37 |
@sonney2k | err area | 17:37 |
blackburn | a minute | 17:37 |
@sonney2k | 0.66? | 17:37 |
blackburn | 0.829 | 17:38 |
blackburn | please note that matplotlib changed axis and the most right is 1.1 | 17:38 |
CIA-90 | shogun: Soeren Sonnenburg master * r7b02400 / (2 files in 2 dirs): whitespace changes - to match coding style - http://bit.ly/hVa4Lk | 17:38 |
blackburn | sonney2k: trapezoid is calculated by halfed (precision_last + precision) multiplied on recall 'shift' | 17:39 |
blackburn | sonney2k: no neg areas | 17:39 |
blackburn | sorry again, said some sh-t :( | 17:40 |
blackburn | neg areas will be here if we will calculate another trapezoids, 'rotated' | 17:40 |
@sonney2k | blackburn, ok then all good | 17:42 |
* sonney2k stops worrying | 17:43 | |
blackburn | sonney2k: yeah, seems all good and I just got myself raveled | 17:44 |
blackburn | sonney2k: seems you didn't made any great decision during the meeting :) | 17:45 |
@sonney2k | blackburn, what did you expect? that we fire everyone ;-) | 17:45 |
@sonney2k | all I wanted was make some official plan and raise awareness that even though the next deadline is far away it is actually closer than thought... | 17:46 |
blackburn | sonney2k: yeah, a good way to start | 17:46 |
@sonney2k | and that every students writes a very brief thingy every week to show he is progressing | 17:47 |
blackburn | sonney2k: to mailing list? | 17:47 |
f-x | sonney2k: any important issues with the patch that i must think about in the immediate future? | 17:50 |
blackburn | sonney2k: hehe, Konrad mixed up sploving with me :D | 17:51 |
@bettyboo | ho | 17:51 |
-!- f-x [~gen@210.212.58.229] has quit [Quit: Work!] | 17:55 | |
-!- Tanmoy [75d35896@gateway/web/freenode/ip.117.211.88.150] has quit [Quit: Page closed] | 19:18 | |
-!- Ziyuan [~Ziyuan@116.21.255.174] has quit [] | 19:31 | |
* sonney2k Re | 20:01 | |
* sonney2k wonders what to do next in shogun | 20:01 | |
blackburn | I know some good idea to implement :D | 20:02 |
@bettyboo | blackburn, hihi | 20:02 |
@sonney2k | blackburn, let me guess - sth about feature selection? | 20:05 |
blackburn | да | 20:05 |
blackburn | *da | 20:05 |
blackburn | :D | 20:05 |
@bettyboo | lolomat | 20:05 |
serialhex | getting your keyboard maps mixed up blackburn? | 20:05 |
blackburn | serialhex: nope :) just joking | 20:06 |
-!- ChanServ changed the topic of #shogun to: Shogun Machine Learning Toolbox | We participate in GSoC with 5 projects http://bit.ly/eI31bL - all hail the code! | Channel logs http://bit.ly/eQDk8Q | 20:06 | |
blackburn | when it is mixed up I say ыщьуешртпт дшлу ерфе | 20:06 |
@sonney2k | da is sth I can understand | 20:06 |
@sonney2k | njet too :) | 20:06 |
blackburn | njet molotoff | 20:07 |
blackburn | njet molotoff | 20:07 |
blackburn | :D | 20:07 |
blackburn | valehtat something-something itche bobrikoff | 20:07 |
serialhex | yschueshrtpt dshlu erfe is what google translate gives me :( | 20:07 |
blackburn | serialhex: it doesnt' make sense to me at all :) | 20:07 |
serialhex | nor me :P | 20:07 |
@sonney2k | serialhex, you know ruby right? | 20:08 |
* sonney2k things of getting a minimal ruby example to run with shogun | 20:08 | |
serialhex | yeah, i'm pretty good with it, why? | 20:08 |
@sonney2k | or alternatively c# | 20:09 |
* serialhex knows no c# | 20:09 | |
* sonney2k compiles shogun w/ ruby support | 20:09 | |
serialhex | except it's related to c/c++ | 20:09 |
serialhex | ..and objective c and (distantly) to java and a bunch of other langs :P | 20:09 |
@sonney2k | blackburn, you do the algorithms - we do the interfaces | 20:09 |
@sonney2k | serialhex, doesn't matter - lets get some minimal ruby to work | 20:10 |
@sonney2k | how do I load a module in ruby? | 20:10 |
serialhex | ok... | 20:10 |
blackburn | :D договорились | 20:10 |
@bettyboo | great | 20:10 |
@sonney2k | and how do I create an object | 20:10 |
serialhex | require 'my_awesome_module' | 20:10 |
@sonney2k | in ' ' ? | 20:10 |
serialhex | a = MyObject.new | 20:10 |
serialhex | typically | 20:10 |
serialhex | yeah | 20:10 |
@sonney2k | without () ? so e.g. if I wanted to create a shogun kernel I could do | 20:11 |
@sonney2k | require 'Kernel' | 20:11 |
serialhex | so if you have a module or library named Awesomeness it would be | 20:11 |
@sonney2k | k=GaussianKernel.new | 20:11 |
@sonney2k | and now if I want to access a member function - I would do k.get_width() ? | 20:11 |
serialhex | yes, though i think we're going to have to rename Kernel in ruby as one of the base objects is named 'Kernel' | 20:11 |
* serialhex goes and checks | 20:11 | |
@sonney2k | and how do I print sth out? | 20:12 |
@sonney2k | like printf in C? | 20:12 |
serialhex | puts 'my awesome stuff' | 20:12 |
@sonney2k | so puts k.get_width() would work? | 20:12 |
* sonney2k creates a ruby_modular folder in examples | 20:12 | |
serialhex | puts will automagically put a newline after each arg, they also have print, printf and simply p which i think all do the same thing as in C | 20:12 |
serialhex | should | 20:12 |
blackburn | my mentor asked me how soon I want to start, I'm confused :) | 20:13 |
serialhex | tell him you'll start yesterday | 20:13 |
@sonney2k | blackburn, well official start is May 23rd IIRC | 20:14 |
blackburn | sonney2k: he meant some another 'start' I think | 20:14 |
@sonney2k | anyathing before that - while very welcome - is up to you | 20:14 |
serialhex | sonney2k: "Kernel" is one of the base modules in ruby, so shogun's Kernel will have to be renamed to something else | 20:14 |
@sonney2k | blackburn, I think he talks about 40hrs/week | 20:14 |
-!- ameerkat [~ameerkat@184-98-140-155.phnx.qwest.net] has quit [Ping timeout: 246 seconds] | 20:14 | |
blackburn | sonney2k: if I will work 40hrs a week I could finish a whole my project during 2-3 weeks | 20:15 |
blackburn | anyway I don't have capability to work 40hrs now ;) | 20:15 |
* serialhex thinks blackburn is an overachiever | 20:15 | |
* sonney2k thinks blackburn should do all the remaining projects that did not get into GSoC in his spare time :D | 20:16 | |
* serialhex agrees with sonney2k!!! | 20:16 | |
blackburn | at first I should pass my exams ;) | 20:16 |
@sonney2k | [20:16][sonne@no:~/Documents/priv/repositories/shogun/src/ruby_modular]$ LD_LIBRARY_PATH=../libshogun irb1.8 | 20:17 |
@sonney2k | irb(main):001:0> require 'Kernel' | 20:17 |
@sonney2k | ./Kernel.so: warning: rb_define_const: invalid name `__STDC_FORMAT_MACROS' for constant | 20:17 |
@sonney2k | => true | 20:17 |
@sonney2k | irb(main):002:0> gk=GaussianKernel.new() | 20:17 |
@sonney2k | => #<Kernel::GaussianKernel:0x7f00d6c50958> | 20:17 |
@sonney2k | irb(main):003:0> puts gk.get_width() | 20:17 |
@sonney2k | 1.0 | 20:17 |
@sonney2k | => nil | 20:17 |
@sonney2k | yikes! works | 20:17 |
-!- sonney2k was kicked from #shogun by bettyboo [flood] | 20:17 | |
-!- sonney2k [~shogun@87.118.92.43] has joined #shogun | 20:17 | |
-!- mode/#shogun [+o sonney2k] by ChanServ | 20:17 | |
blackburn | bad betty | 20:17 |
* sonney2k smacks betty | 20:17 | |
@bettyboo | blackburn: great you are back, betty | 20:17 |
serialhex | lol... bettyboo dosnt like you much sonney2k!! :P | 20:17 |
blackburn | :D | 20:17 |
@bettyboo | serialhex: the boosting-like algorithm for variable-dimensinaol features | 20:17 |
serialhex | yeah, ruby is pretty sweet like that... you can do almost anything out of the box | 20:18 |
@sonney2k | serialhex, anyway works - let me commit that - so only typemaps are missing now | 20:18 |
blackburn | why we have only KRR and SVR for regression? | 20:18 |
serialhex | http://tryruby.org/ if you want to play with it & have a little tutorial | 20:18 |
serialhex | typemaps and examples sonney2k | 20:19 |
@sonney2k | blackburn, good question - I very much (and I know cwidmer too) would love to see GPs in there too | 20:19 |
@sonney2k | or whatever | 20:19 |
@sonney2k | serialhex, yes but examples are an easy port | 20:19 |
blackburn | anyway I'm not acknowledged with GPs ;) | 20:19 |
@sonney2k | serialhex, syntax is 99% the same just data loading not | 20:19 |
@sonney2k | blackburn, what does that mean? | 20:20 |
blackburn | sonney2k: I never implemented GPs in any language at all | 20:20 |
blackburn | even didn't sure I know what it is :D | 20:20 |
@bettyboo | 8) | 20:20 |
@sonney2k | blackburn, me neither - and I've never used them... | 20:20 |
serialhex | a Q about that sonney2k, i've found only 2 libraries that deal with numeric things in ruby: NArray & ruby bindings to the GNU Scientific Lbrary... | 20:21 |
CIA-90 | shogun: Soeren Sonnenburg master * rd83b31e / examples/undocumented/ruby_modular/minimal_test.rb : add ruby modular minimal example - http://bit.ly/mr76DK | 20:21 |
CIA-90 | shogun: Soeren Sonnenburg master * rbbadc47 / .gitignore : ignore *.class files too - http://bit.ly/j1Leaf | 20:21 |
CIA-90 | shogun: Soeren Sonnenburg master * r5dbbbb2 / (2 files): rename variables to fix naming clashes - http://bit.ly/m4NC7L | 20:22 |
serialhex | NArray is self-contained but they say it might be changed at a later date | 20:22 |
@sonney2k | serialhex, I guess ruby has arrays too? | 20:22 |
@sonney2k | or lists? | 20:22 |
serialhex | and the GSL bindings requires you to have GSL installed but seems more mature | 20:22 |
serialhex | sonney2k: arrays, hashes... | 20:22 |
@sonney2k | it would make sense to write typemaps for lists of strings and also for arrays ( aka vectors) | 20:22 |
@sonney2k | for a starter | 20:22 |
serialhex | a = ['foo', 'bar','baz'] | 20:22 |
@sonney2k | then matrix / and or sparse matrices | 20:22 |
@sonney2k | serialhex also a = [1,2,3] ? | 20:23 |
serialhex | yeah | 20:23 |
serialhex | or | 20:23 |
@sonney2k | yeah then for that... | 20:23 |
* sonney2k will switch to c# | 20:23 | |
serialhex | so as far as the typemaps go i kind of have 3 opts, NArray, GSL or Shogun | 20:23 |
@sonney2k | serialhex, yeah I don't know which one is good... you could of course do typemaps for all of them | 20:26 |
serialhex | ok... i'll see what i can do with plain shogun, and then work on the others | 20:27 |
serialhex | that may not be the easiest way, so i may change, but we'll see... | 20:27 |
* blackburn thinks serialhex should be in GSoC students :D | 20:28 | |
* sonney2k all I can promise is to properly acknowlegde you... | 20:29 | |
@sonney2k | ^^ serial | 20:29 |
serialhex | blackburn, sploving beat me by 3 languages... what can i do but concede it to him?? | 20:29 |
blackburn | serialhex: beat him physically | 20:30 |
serialhex | np sonney2k, i'll get back to you on that Q i asked you a while ago also... i havnt had a chance to talk to her | 20:30 |
@sonney2k | serialhex, and he has contributed a new language to swig... | 20:30 |
serialhex | lol blackburn | 20:30 |
@bettyboo | :^) | 20:30 |
@sonney2k | serialhex, I asked my boss but no news yet (bosses are always busy) | 20:30 |
serialhex | i know :( | 20:31 |
blackburn | I also want to be a busy boss | 20:31 |
blackburn | serialhex: obey me! | 20:31 |
serialhex | btw, i like the channel message: 'all hail the code!' | 20:31 |
* sonney2k blackburn obey me and goo code! | 20:31 | |
serialhex | goo code??? :P | 20:31 |
@sonney2k | go | 20:32 |
blackburn | sonney2k: ich bin muede | 20:32 |
@sonney2k | serialhex, you've seen futurama and the hypnotoad? | 20:32 |
@sonney2k | I am a big fan of it | 20:32 |
@sonney2k | http://futurama.wikia.com/wiki/Hypnotoad | 20:32 |
serialhex | i've seen a bit of futurama, but i dont recall a hypnotoad | 20:32 |
blackburn | btw guys do you know what is on my gravatar on github? | 20:33 |
serialhex | i havnt a clue blackburn | 20:33 |
blackburn | sonney2k: what about you? | 20:33 |
serialhex | cool cool | 20:34 |
@sonney2k | blackburn, some paper bag with glasses? | 20:34 |
blackburn | it is from 12 oz mouse | 20:34 |
@sonney2k | the bag seems a little depressed and not feeling well judging from face color | 20:35 |
serialhex | though i find this works better: http://xkcd.com/149/ | 20:35 |
serialhex | i agree with sonney2k on the depressed thing | 20:35 |
blackburn | his name is 'square guy' :D | 20:35 |
@sonney2k | blackburn, fits your vodka addicition ;-) | 20:36 |
blackburn | oh yeah | 20:36 |
* sonney2k tries to figure out how to compile a c# program | 20:36 | |
* serialhex hands sonney2k a magical microsoft wand and tells him to be careful, it could make his computer BSOD | 20:37 | |
@sonney2k | not touching m$ thingies except their keyboards | 20:38 |
@sonney2k | I will try some mono tutorial | 20:39 |
serialhex | i agree, i i have win 7 installed on my computer for the job i had, but in the process of getting my computer set up for GSoC the bootloader got messed up and now it dosnt boot... that was like a month ago and havnt had a reason to fix it yet | 20:40 |
@sonney2k | I only boot windows for doing my tax declaration | 20:41 |
blackburn | as for me I boot windows for play microsoft flight simulator a little :D | 20:42 |
serialhex | yeah, thats about all windoze is good for, taxes & games | 20:43 |
blackburn | I don't have to pay taxes for it is not the case for me :D | 20:43 |
blackburn | for it^S | 20:43 |
serialhex | lucky you! | 20:45 |
blackburn | will have to a bit later | 20:46 |
@sonney2k | so the mono stuff compiled | 20:48 |
@sonney2k | and a mono hello world works too | 20:48 |
@sonney2k | now for the shogun mono thingy | 20:48 |
blackburn | How to learn many things during wednesday evening with... | 20:49 |
blackburn | Soeren Sonnenburg! | 20:49 |
@sonney2k | seems like it is the same as with java | 20:50 |
blackburn | if I will be german I could be not Lisitsyn but Fuchs :D | 20:51 |
blackburn | may be 'if I would' would be better :) | 20:53 |
blackburn | serialhex: seems for you my writings looks like tajikistan man speaks russian | 20:54 |
serialhex | blackburn you are correct! though 'if i was german i would not be Lisitsyn but Fuchs' would be best :D | 20:54 |
blackburn | oh! yeah, I recall some english lessons :) | 20:55 |
* serialhex is 0.5 lost on blackburn's last comment :P | 20:55 | |
serialhex | we need to get you better (or more) lessons :P | 20:55 |
blackburn | serialhex: when some asian man speaks russian it looks the same | 20:55 |
serialhex | ahh, ok, now it's much clearer (and what i thought) | 20:56 |
* serialhex imagines how bad my russian would be if i tried :D | 20:56 | |
blackburn | serialhex: we will disappear soon ;) so I have to learn english | 20:56 |
@bettyboo | :*) | 20:56 |
blackburn | we = russians, etc :) | 20:57 |
blackburn | have to go now, will be later | 20:57 |
serialhex | sonney2k, you might be interested in this to help with organizing everything: https://github.com/nvie/gitflow | 20:58 |
serialhex | it's basically a way to organize a project better with different branches & stuff, pretty interesting & is based on git | 21:00 |
@sonney2k | sigh sth still doesn't work | 21:04 |
@sonney2k | ../../examples/undocumented/csharp_modular/minimal_test.cs(2,7): error CS0246: The type or namespace name `Kernel' could not be found. Are you missing a using directive or an assembly reference? | 21:05 |
@sonney2k | yeah whatever | 21:05 |
serialhex | hmm... idk wtf is wrong | 21:06 |
@sonney2k | hmmhh lots of errors but at least I know now how it could work... | 21:38 |
@sonney2k | hah down to 8 errors :) | 21:41 |
blackburn | here I am! | 21:56 |
blackburn | what's up? C# isn't working? | 21:57 |
serialhex | c# is the devils child!!! | 22:03 |
@sonney2k | blackburn, down to 6 errors now | 22:03 |
serialhex | lol | 22:04 |
blackburn | sonney2k: go Soeren, go! | 22:04 |
* sonney2k forgot to enable compile flags | 22:04 | |
@sonney2k | err debug | 22:04 |
@sonney2k | wasting too much time.... | 22:04 |
blackburn | yeah you have less time with every minute ;) | 22:05 |
@sonney2k | yeah I guess I should have a count down too | 22:06 |
blackburn | 31 forks so far | 22:11 |
blackburn | hehe | 22:11 |
blackburn | there was ~20 a week ago | 22:11 |
-!- f-x [~gen@180.149.49.227] has joined #shogun | 22:12 | |
@sonney2k | zero erros | 22:13 |
@sonney2k | errors | 22:13 |
blackburn | congrats ;) | 22:13 |
serialhex | sweet!!!! | 22:13 |
@sonney2k | but I cannot find the init_shogun function | 22:13 |
@sonney2k | so the example fails... | 22:13 |
blackburn | sonney2k: is it the 'best' period of shogun development? | 22:13 |
blackburn | as for activity, etc | 22:14 |
@sonney2k | lets hope it will be | 22:14 |
@sonney2k | or not hope - maybe it will even get better when you are the mentor | 22:14 |
blackburn | me? | 22:14 |
blackburn | not sure understood you ;) | 22:16 |
serialhex | yeah blackburn, you would be a great mentor!! | 22:17 |
serialhex | "what do you mean you only submitted 3 patches in a month!!! when i was a student i sumbitted 82 patches and now i do 375 patches a month!!!" :P | 22:18 |
blackburn | serialhex: hope I could be a student say twice ;) | 22:19 |
serialhex | you probably can | 22:19 |
blackburn | but next year I will have ehhh | 22:20 |
blackburn | 'defence'? | 22:20 |
@sonney2k | strike! | 22:21 |
@sonney2k | it works | 22:21 |
blackburn | serialhex: how should I say that :D | 22:21 |
blackburn | bachelor's degree qualification work.. defence? | 22:21 |
blackburn | looks like translate word-by-word from russian :) | 22:22 |
blackburn | *even though all my sentences are looking this way | 22:22 |
CIA-90 | shogun: Soeren Sonnenburg master * rbc536fc / (5 files in 3 dirs): Get a minimal csharp example to work. - http://bit.ly/m71znR | 22:28 |
CIA-90 | shogun: Soeren Sonnenburg master * r61fc777 / src/configure : add lib prefix like in java to generated .so files - http://bit.ly/ih6EEp | 22:28 |
@sonney2k | blackburn, as long as we still understand you :) | 22:29 |
@sonney2k | blackburn, so you will have your bachelor or master next year? | 22:29 |
@sonney2k | or PhD? | 22:29 |
blackburn | sonney2k: bachelor | 22:30 |
@sonney2k | sploving, could you add your minimal lua example too? | 22:30 |
@sonney2k | blackburn, so you can be student 2 more times :D | 22:30 |
@sonney2k | or until you have your phd.... | 22:31 |
blackburn | sonney2k: yeap | 22:31 |
blackburn | sonney2k: just the matter of 'fair play', if I was developing shogun for two or more years, other candidates could say it is not fair :) | 22:32 |
blackburn | serialhex: did I said it right way ^? now worrying with my english sentences | 22:34 |
@sonney2k | yeah, if you fail this summer then it would not be fair :P | 22:35 |
serialhex | blackburn: on the phone... i'll let you know in a few | 22:35 |
blackburn | sonney2k: phew! do you really think I could fail? | 22:35 |
@sonney2k | blackburn, no | 22:36 |
@sonney2k | but you were talking about fairness - not me | 22:36 |
* sonney2k considers it very fair if someone spends lots of time developing on shogun to be in GSoC | 22:37 | |
serialhex | blackburn, this sentence: "bachelor's degree qualification work.. defence?" | 22:38 |
serialhex | are you talking about defending a thesis or something blackburn? | 22:40 |
serialhex | (just want to get it right so i can help you better) | 22:40 |
CIA-90 | shogun: Soeren Sonnenburg master * r2e28298 / (2 files in 2 dirs): Merge git://github.com/chengsoonong/shogun - http://bit.ly/lmkF0H | 22:40 |
blackburn | serialhex: yeah, seems to be 'thesis' | 22:41 |
blackburn | sonney2k: I see :) | 22:42 |
serialhex | then i would say "i must defend my bachelor's thesis against the evil & tyrranical thesis-squashers, who will do everything in their power to see me fail!!!" | 22:42 |
serialhex | ...or something to that extent :D | 22:43 |
blackburn | oh thank you! :D | 22:43 |
CIA-90 | shogun: Soeren Sonnenburg master * r9ce80cb / : Merge git://github.com/frx/shogun - http://bit.ly/ikqivV | 22:43 |
serialhex | maybe we can set up skype or something and i could _hear_ your english :P | 22:44 |
@bettyboo | great! | 22:44 |
serialhex | see!!! even she likes it!!! | 22:44 |
blackburn | serialhex: it will blow your mind | 22:44 |
serialhex | i imagine it will | 22:44 |
* sonney2k types ./configure --interfaces=libshogun,libshogunui,python,r,octave,cmdline,octave_modular,r_modular,lua_modular,java_modular,csharp_modular,java_modular,matlab | 22:45 | |
blackburn | sonney2k: it shall compile sloooooow | 22:46 |
@sonney2k | that looks like 400% overkill | 22:46 |
@sonney2k | blackburn, thanks for reminding | 22:46 |
* sonney2k adds --enable-debug --disable-optimizations | 22:46 | |
@sonney2k | s/optimizations/optimization/ | 22:46 |
blackburn | serialhex: why one can't say 'advices'? | 22:47 |
@sonney2k | I guess that will take 30 minutes on my notebook to compile | 22:47 |
blackburn | serialhex: but 'advice' | 22:49 |
blackburn | no plural I mean | 22:49 |
serialhex | blackburn, 'advice' is one of those words that is both singular and plural. so i can say: 'he gave me a bit of advice' or 'he gave me lots of advice' and either is ok' | 22:49 |
blackburn | hehe see now | 22:49 |
serialhex | sometimes things are wonky in english like that, like the plural of ox is oxen, but the plural of box is not boxen :D | 22:50 |
blackburn | serialhex: I was wondering of plural of axis | 22:50 |
serialhex | it's boxes... and there are a bunch of other stupid rules that i can probably find | 22:50 |
serialhex | like coordinate axis? | 22:51 |
blackburn | yeah, it is axes iirc | 22:51 |
blackburn | just like weapon, right? | 22:51 |
serialhex | no, i dont think so | 22:52 |
* serialhex is checking for proper use & spelling | 22:52 | |
blackburn | it is! pl of axis is axes | 22:53 |
blackburn | I googled ;) | 22:53 |
serialhex | yes you are right! | 22:53 |
serialhex | i wiktionaried :P | 22:53 |
@bettyboo | haha | 22:54 |
serialhex | though one would pronounce it like 'axees' and the plural of ax like 'axis' (or more like 'ackzes' and 'ackzis' you will probably have to hear it to get it right though) | 22:55 |
blackburn | ah I see | 22:56 |
serialhex | yeah: http://en.wiktionary.org/wiki/axes#English | 22:56 |
blackburn | i <-> e | 22:57 |
serialhex | ??? | 22:57 |
blackburn | in pronunciation | 22:57 |
blackburn | one have i, another sth like e :) | 22:58 |
blackburn | sonney2k: how it's going? compiling? | 22:59 |
@sonney2k | blackburn, found some new bugs ... trying to fix them | 23:00 |
@sonney2k | protected / public issues that I introdcued in the csharp thing | 23:00 |
@sonney2k | (only with static interfaces) | 23:00 |
blackburn | sonney2k: how you mind some image processing examples in shogun? may be I could work with it in future | 23:02 |
blackburn | image classifying etc | 23:02 |
@sonney2k | blackburn, not at all | 23:02 |
blackburn | ok | 23:02 |
serialhex | ooh, blackburn your going to have to hear it because i cant really explain the 'sound' of it to you... i've seen some english dictionaries that have an option to speak a word but i cant recall any off the top of my head :( | 23:03 |
blackburn | serialhex: I think I probably understand the sound | 23:04 |
serialhex | ok good | 23:04 |
blackburn | sonney2k: but why we need some real-valued things if shogun is working with dna strings, etc? | 23:05 |
@sonney2k | blackburn, ? | 23:05 |
blackburn | for example are my proposed algorithms suitable to your research tasks? | 23:06 |
blackburn | I'm just interested with 'roadmap' of shogun | 23:07 |
@sonney2k | blackburn, I am not only doing bioinformatics | 23:09 |
@sonney2k | also image stuff, computational chemistry etc | 23:09 |
blackburn | image stuff? but why you are against it in shogun? | 23:10 |
@sonney2k | blackburn, ahh I said 'not at all' - which means I am all for it | 23:10 |
@sonney2k | I don't have anything against it in the context above | 23:11 |
blackburn | ooh | 23:11 |
@sonney2k | dangerous misunderstandings.... | 23:11 |
blackburn | do not worry it is because I'm stupid :D | 23:11 |
-!- serialhex-10 [~androirc@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 23:14 | |
blackburn | sonney2k: the only reason why I am asking for this prospects - I probably will have image-processing related bachelor's thesis (cause 95% of my teachers is working in that field) | 23:14 |
blackburn | and I think linking it with shogun could be nice | 23:15 |
serialhex-10 | hey sonney2k, chanserv says you are still looking for stuents | 23:15 |
@sonney2k | blackburn, I don't mind - you should however also look at opencv it has lots of image stuff - maybe you can connect the projects | 23:15 |
@sonney2k | serialhex, someone has to fix this then | 23:16 |
serialhex-10 | thats kind of spiffy blackburn | 23:16 |
blackburn | sonney2k: ok, anyway it will be later | 23:16 |
blackburn | much later | 23:16 |
@sonney2k | opencv has 12 students this year... | 23:17 |
blackburn | I don't really like image processing though | 23:17 |
blackburn | but seems I have no choice :D | 23:17 |
@bettyboo | interesting | 23:18 |
@sonney2k | I have a nice example here where we detect from the face if a person is male or female - so it can be fun :) | 23:18 |
blackburn | sonney2k: I had an example of face/non-face recoginition in my set of lectures | 23:18 |
blackburn | I have used kNN for it :D just for example | 23:19 |
@sonney2k | but that is not as cool as predicting gender :) | 23:19 |
serialhex | yeah, i like sonney2k's example... that's pretty damn spiffy | 23:19 |
blackburn | exactly | 23:20 |
blackburn | just mentioned :) | 23:20 |
* serialhex wonders if it can predict the gender of transvestites/transexuals... :P | 23:20 | |
@sonney2k | http://sonnenburgs.de/soeren/talks/2010-06-23-COFFIN-ICML.pdf | 23:20 |
@sonney2k | look at the talk I gave | 23:20 |
serialhex | the _true_ gender :D | 23:20 |
@sonney2k | and the predoictions for the kelly family at the end :) | 23:21 |
@sonney2k | predictions | 23:21 |
blackburn | sonney2k: really cool! | 23:21 |
-!- serialhex-10 [~androirc@99-101-149-136.lightspeed.wepbfl.sbcglobal.net] has quit [Read error: Operation timed out] | 23:21 | |
@sonney2k | even managed to predict the gender of our prime minister correctly | 23:22 |
serialhex | sonney2k: is bettyboo using shogun?? or do you not know? | 23:23 |
@bettyboo | serialhex: also, there are no classes that provide general optimization methods, right? or did I miss something again? :) | 23:23 |
CIA-90 | shogun: Soeren Sonnenburg master * r7b926a5 / (2 files in 2 dirs): | 23:23 |
CIA-90 | shogun: fix compile failure | 23:23 |
CIA-90 | shogun: It got introduced in the csharp commit (protected variables | 23:23 |
CIA-90 | shogun: where accessed - introducing public getters fixes this). - http://bit.ly/msA3pu | 23:23 |
CIA-90 | shogun: Soeren Sonnenburg master * r4d0fe5d / (2 files): rename functions to reflect that we obtain the global objects - http://bit.ly/l1Eixd | 23:23 |
@sonney2k | blackburn, hmmhh r_modular does not compile with evaluations | 23:23 |
@sonney2k | anyway have to sleep now | 23:23 |
@sonney2k | cu all! | 23:23 |
blackburn | damn! | 23:23 |
blackburn | what is the problem? | 23:23 |
blackburn | ah nevermind | 23:24 |
blackburn | just let me know what is the problem or fix it :) | 23:24 |
blackburn | serialhex: остались только ты и я, это так романтично | 23:24 |
* serialhex google-translating :D | 23:25 | |
serialhex | romantic?? | 23:25 |
blackburn | :D | 23:25 |
serialhex | thats kind of weird dude :P | 23:25 |
blackburn | oh gay jokes on russian is kinda weird right | 23:26 |
* serialhex thinks typemaps are a pain in the ass!! | 23:26 | |
serialhex | heh, gay jokes in general are kind of weird, but i'm secure so it's ok | 23:27 |
blackburn | ok sweety | 23:27 |
serialhex | though you shouldnt be talking about us out loud hun! it might give everyone the wrong impression!! | 23:28 |
blackburn | sorry sweety :D | 23:29 |
serialhex | for some reason google translate always defaults to czech for me... idfk why or how to change it :-/ | 23:30 |
blackburn | serialhex: may be you know czech | 23:30 |
blackburn | some prznzn phrzssh | 23:30 |
serialhex | google translate gives me: prznzn phrzssh (and says it's german :P ) | 23:32 |
blackburn | I just wrote some prazska prazksksk sshhskks | 23:33 |
serialhex | it says it can't translate bosnian to english yet | 23:34 |
blackburn | :D | 23:34 |
@bettyboo | 8) | 23:34 |
* serialhex is still trying to learn the rest of ruby, and is not ready for russian :D | 23:36 | |
serialhex | or german or bosnian or fsck-nian :P | 23:36 |
serialhex | blackburn: have you seen/read this (or anything on it) http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1000211 i think this would be cool to implement in shogun if someone had the time/skill | 23:54 |
blackburn | serialhex: don't know what it is | 23:56 |
serialhex | ahh, it's a pretty interesting theory & machine learning thing that is kind of like how various cortical columns in the brain communicate to each other... via a heirarchy | 23:57 |
serialhex | it uses bayesian learning & a 'free energy bound' on suprise (essentially what it dosnt know) | 23:58 |
* serialhex wonders if shogun has any bayesian learning algorithms | 23:59 | |
--- Log closed Thu Apr 28 00:00:01 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!