verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
DMS has joined #mlpack
< DMS>
Hey guyz
< DMS>
I'm having Ubuntu 16.04
< DMS>
Tell me the best source for installing mlpack
< rcurtin>
DMS: you can just use apt
< rcurtin>
sudo apt-get install libmlpack-dev
< rcurtin>
that should be the easiest thing :)
< DMS>
yeah thatz fine
< DMS>
and what about building mlpack
< DMS>
i'm following the steps given in readme file
< DMS>
it's ok rite?
< rcurtin>
yeah... the readme contains correct instructions
< DMS>
what about armadillo
< DMS>
can you give me the command to install it ?
< DMS>
but finally after configuring cmake and all
< DMS>
i entered command make
< DMS>
and i got an error , I'm unable to figure it out
< DMS>
please help me
< nilay>
can you post the error
< nilay>
did you follow the instructions, make a build dir and write cmake ..
< DMS>
In file included from /home/mohith/Desktop/mlpack-2.0.1/src/mlpack/../mlpack/methods/kmeans/kmeans.hpp:199:0, from /home/mohith/Desktop/mlpack-2.0.1/src/mlpack/methods/gmm/em_fit.hpp:22, from /home/mohith/Desktop/mlpack-2.0.1/src/mlpack/methods/gmm/gmm.hpp:22, from /home/mohith/Desktop/mlpack-2.0.1/src/mlpack/methods/gmm/gmm.cpp:16: /home/mohith/Desktop/mlpack-2.0.1/src/mlpack/../mlpack/me
< nilay>
are you sure you are pasting the error and not the warning? paste the line which contains the word "error:"
< DMS>
error: there are no arguments to ‘isnan’ that depend on a template parameter, so a declaration of ‘isnan’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm))
< DMS>
error: there are no arguments to ‘isinf’ that depend on a template parameter, so a declaration of ‘isinf’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm)) ^ src/mlpack/CMakeFiles/mlpack.dir/build.make:686: recipe for target 'src/mlpack/CMakeFiles/mlpack.dir/methods/gmm/gmm.cpp.o' failed make[2]: *** [src/mlpack/CMakeFiles/mlpack.dir/methods/gmm/gmm.cpp.o] Error 1 CMakeFiles
DMS has quit [Ping timeout: 250 seconds]
DMS_ has joined #mlpack
< DMS_>
hey bro did you figure out my error?
< DMS_>
what was the problem
< DMS_>
error: there are no arguments to ‘isnan’ that depend on a template parameter, so a declaration of ‘isnan’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm)) ^
< DMS_>
error: there are no arguments to ‘isinf’ that depend on a template parameter, so a declaration of ‘isinf’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm)) ^ src/mlpack/CMakeFiles/mlpack.dir/build.make:686: recipe for target 'src/mlpack/CMakeFiles/mlpack.dir/methods/gmm/gmm.cpp.o' failed make[2]: *** [src/mlpack/CMakeFiles/mlpack.dir/methods/gmm/gmm.cpp.o] Error 1
< nilay>
can you tell me the steps that you did. this should not occur if cmake generates makefile correctly
< DMS_>
i installed all those three - boost , armadillo , cmake
< DMS_>
then i used commands given in read me
< DMS_>
cmake ../
< DMS_>
cmake -D DEBUG=ON -D PROFILE=ON ../
< DMS_>
DEBUG=(ON/OFF)
< DMS_>
PROFILE=(ON/OFF)
< DMS_>
ARMA_EXTRA_DEBUG=(ON/OFF)
< nilay>
did you make a new directory and change to that directory?
< DMS_>
yeah
< nilay>
ok
< DMS_>
i named it build
< DMS_>
i did that first before i installed anything
< nilay>
are there other errors also that you are having? can you post the errors here : http://pastebin.com/
< nilay>
i think there is some problem with files not being correctly included
< DMS_>
which files
< nilay>
you can try doing the whole process again, or wait here until someone can guide you.
< DMS_>
oh
< DMS_>
should we set some path or anything like that?
< nilay>
make a new directory build2 and do cmake .. in it
< nilay>
if same error comes then i guess wait, someone will solve your problem
< DMS_>
those commands given in read me are fine rite
< DMS_>
at one step ... they gave some thing like /configure and i didn't follow that bcoz i didn't understand
< DMS_>
instead i used command given below cmake ../
< DMS_>
is there any mistake?
< DMS_>
hey bo
< DMS_>
after entering command cmake ../
< DMS_>
i got some warnings
< DMS_>
CMake Warning (dev) at src/mlpack/CMakeLists.txt:84 (add_custom_target): Policy CMP0037 is not set: Target names should not be reserved and should match a validity pattern. Run "cmake --help-policy CMP0037" for policy details. Use the cmake_policy command to set the policy and suppress this warning.
< DMS_>
The target name "test" is reserved or not valid for certain CMake features, such as generator expressions, and may result in undefined behavior. This warning is for project developers. Use -Wno-dev to suppress it.
< DMS_>
Could NOT find Doxygen (missing: DOXYGEN_EXECUTABLE) CMake Warning at CMakeLists.txt:331 (message): txt2man not found; man pages will not be generated.
< nilay>
and now when you do make
DMS_ has quit [Ping timeout: 250 seconds]
DMS has joined #mlpack
< DMS>
hey after entering cmake ../
< DMS>
CMake Warning (dev) at src/mlpack/CMakeLists.txt:84 (add_custom_target): Policy CMP0037 is not set: Target names should not be reserved and should match a validity pattern. Run "cmake --help-policy CMP0037" for policy details. Use the cmake_policy command to set the policy and suppress this warning.
< DMS>
Could NOT find Git (missing: GIT_EXECUTABLE) CMake Warning (dev) at src/mlpack/CMakeLists.txt:84 (add_custom_target): Policy CMP0037 is not set: Target names should not be reserved and should match a validity pattern. Run "cmake --help-policy CMP0037" for policy details. Use the cmake_policy command to set the policy and suppress this warning.
< nilay>
don't worry about those warnings
< nilay>
do make again
< DMS>
The target name "test" is reserved or not valid for certain CMake features, such as generator expressions, and may result in undefined behavior. This warning is for project developers. Use -Wno-dev to suppress it. -- Could NOT find Doxygen (missing: DOXYGEN_EXECUTABLE) CMake Warning at CMakeLists.txt:331 (message): txt2man not found; man pages will not be generated.
< DMS>
are you sure bro?
< nilay>
yes
< nilay>
i also get these warnings
< DMS>
oh
< DMS>
again same error bro
< DMS>
whatz the problem?
< DMS>
i just followed all the commands given in read me
< DMS>
should I do anything other than that?
< nilay>
ok so are you sure you are posting all the errors, can you scroll above and see if there are more
< mohith>
error: there are no arguments to ‘isnan’ that depend on a template parameter, so a declaration of ‘isnan’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm))
< mohith>
what is this error?
< mohith>
hey bro anybody free ?
< mohith>
Please help me I'm screwed
nilay has joined #mlpack
< mohith>
hey nilay
< mohith>
help me man
< mohith>
i got screwed up
< mohith>
error: there are no arguments to ‘isnan’ that depend on a template parameter, so a declaration of ‘isnan’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm))
< mohith>
what is this error?
< nilay>
mohith: wait for some time, you'll get help don't worry
< mohith>
okay
mohith has quit [Ping timeout: 250 seconds]
< nilay>
rcurtin: we need to put license header information in all the files in arma_extend ?
< nilay>
i can see commits to "remove extraneous licensing information"
Mathnerd314 has quit [Ping timeout: 246 seconds]
mentekid has joined #mlpack
nilay has quit [Quit: Page closed]
nilay has joined #mlpack
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#1036 (master - 779556f : sumedhghaisas): The build was broken.
< nilay>
zoq: the discretize function that i added currently takes a lot of time (31s for 2 images present in the example folder), can we do something about this, especially when we are ignoring most of the data (20,000 * 256) labels we have and we only take one label out of these, that is the most "representative" and returning it..
< rcurtin>
mohith: why do you need to compile mlpack by hand? why not just install through apt?
< rcurtin>
anyway if you really do want to compile by hand, I think that you should use the git master branch
< rcurtin>
oh, I didn't see that he left the channel, oops...
< rcurtin>
nilay: you're right, we should have put license header information in, I took care of it (you can see #702 for more info if you like)
< rcurtin>
sorry about that, I should have realized it
< nilay>
rcurtin: thanks for fixing it. i didn't knew that open source communities also take license so seriously.
< rcurtin>
yeah, I'm not sure what the situation is like for many open source projects but I know Conrad takes licenses pretty seriously, he has done a lot of research and is quite knowledgeable about IP law with respect to open source software licenses
< zoq>
nilay: I'm in a meeting, so it might take some time before I can take a look at the performance issue.
< nilay>
rcurtin: yeah i guess armadillo also has a commercial license, so maybe that is why there are some exceptions for some projects.
< nilay>
zoq: no problem
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#1038 (master - 20b4b27 : Ryan Curtin): The build was fixed.
< nilay>
rcurtin: to backport functions which are member of a class in armadillo, what different do i have to do?
< nilay>
method of a class*
< nilay>
or is it not a good idea to do this. i was planning to backport index_min() and index_max() that were added in arma 7.2 , but i only use index_min() once in my code, uptil now, so i can also write a trivial function for that.
dms has joined #mlpack
< dms>
error: there are no arguments to ‘isnan’ that depend on a template parameter, so a declaration of ‘isnan’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm))
< dms>
hello ! can anyone tell me how to resolve this error ?
< dms>
error: there are no arguments to ‘isinf’ that depend on a template parameter, so a declaration of ‘isinf’ must be available [-fpermissive] if (isnan(cNorm) || isinf(cNorm))
< dms>
I started with mlpack-2.0.2 , the link you gave me above
< dms>
and tell me one thing ..is it mandatory to do all those -D flag commands ?
< dms>
It is not neccessary rite?
< zoq>
dms: If you use the stable version it should be sufficient to just call 'cmake ..' and 'make'
< dms>
okay ...
< zoq>
If you use the master branch it's probably a good idea to set '-D DEBUG=ON -D PROFILE=ON'
mentekid has quit [Ping timeout: 260 seconds]
< zoq>
nilay: There is no need to center the data before calling PCA, we do that already with: math::Center(data, centeredData); in pca.cpp.
< zoq>
nilay: Also, I don't think you have to convert the result by using: arma::conv_to<arma::vec>::from(...). Also, we could use Keon's recently introduced binarize function that uses OpenMP to speed things up.
< zoq>
nilay: Did you build with DEBUG=OFF and PROFILE=OFF?
mentekid has joined #mlpack
< zoq>
nilay: I'm not sure, but using 'arma::mat zs = labels.cols(z1) == labels.cols(z2)' could speed things up
mentekid has quit [Read error: Connection reset by peer]
< nilay>
zoq: yes i normally build with these things off
< nilay>
zoq: i think the slowest step in all this is calling the PCA function for 256*20000 label
< zoq>
nilay: maybe you can time that particular method call by using mlpack's timer functionality?
< nilay>
and yeah thanks for the rest of the suggestions. what do i ignore when i compile with off setting.
< nilay>
yeah i can
< nilay>
give me a minute, i'll post the time here
< zoq>
Most of all you build with compiler optimizations turned on '-O3'
< nilay>
also do you think, that writing our own small function for index_min is better instead of backporting arma
< nilay>
i can use openmp to speed up that small function later
< nilay>
backporting from arma*
< zoq>
I don't think it's necessary, to backport the function.
< nilay>
what command you use for cmake ..
< zoq>
'cmake -D DEBUG=OFF -D PROFILE=OFF ..'
< zoq>
dms: I meant to use '-D DEBUG=OFF -D PROFILE=OFF' when using the master branch.
< zoq>
no, just wanted to point out that if you use the master branch it's '-D DEBUG=OFF -D PROFILE=OFF' and not '-D DEBUG=ON -D PROFILE=ON'
< dms>
okay fine
< dms>
and final thing should why are we doing make install again ?
< dms>
should i do that ... i should change root permissions rite again ?
< dms>
doing that is recommended or what?
< nilay>
what about obmake
< nilay>
oops sorry
< zoq>
dms: to move headers, libs, executables into the default locations for headers and libs ... you don't necessarily have to do 'make install' , you can also use everything from within the build folder
mentekid has joined #mlpack
< dms>
so i've successfully built mlpack in my pc
< dms>
so should i start with documentation or tutorials or anything else
< zoq>
dms: great, starting with the tutorials is a good idea
< dms>
okay thank you zoq..:D
< zoq>
dms: you're welcome, glad I could help
< nilay>
i am giving -verbose while running the executable it doesn't show the time
mentekid has quit [Ping timeout: 246 seconds]
< dms>
hey zoq ...while implementing any algorithm ...where should i have the data
sumedhghaisas has joined #mlpack
< dms>
is it in bin folder ?? where we have all the executable s
< zoq>
nilay: you mean the edge_boxes executable not the unit test right?
< zoq>
dms: you could use src/mlpack/tests/data
nilay has quit [Ping timeout: 250 seconds]
dms has quit [Ping timeout: 250 seconds]
dms_ has joined #mlpack
< dms_>
hey zoq I'm little bit confused ...I'm planning to implement linear regression since it is my first learning algorithm
< dms_>
tell me how to do that ...
< dms_>
where to include data and all ..!!
nilay has joined #mlpack
< nilay>
ok so, discretize function takes a total of 27.72 seconds, of which 8.75 seconds is for pca and rest for other things.
< nilay>
feature extraction takes 15.74 seconds
< nilay>
these times are for only images in example folder
< zoq>
nilay: oh, okay, can you post the code you used to time everything? I'll take a look at it later today.
< zoq>
nilay: thanks, that makes things easier, also thanks for pushing the mapping code. I'll try to incorporate the code as fast as possible. The plan for this week is to work on the inception model right?
< nilay>
zoq: yes
< zoq>
nilay: okay, great :)
nilay has quit [Ping timeout: 250 seconds]
nilay has joined #mlpack
< nilay>
zoq: so what do you think after reading these things, my first step should be
< nilay>
also the interface that we discussed for edge_boxes, a similar type of discussion here could help greatly.
dms has joined #mlpack
< dms>
how to become familiar to the libraries ? Where to start ? I know ML algorithms...I've already implemented most of them in matlab. What to do know? Should i implement them using mlpack?If yes how?
< rcurtin>
dms: a lot of the algorithms you are familiar with are probably already implemented in mlpack, so taking a look through the source code is probably a good idea
< dms>
can you tell me how can we run programs using simple commands ?
< dms>
how about giving input files as input and getting output file??
< rcurtin>
dms: have you read any of the tutorials or guides? or tried to run any of the command-line programs with the "--help
< rcurtin>
" option?
< zoq>
nilay: I think the first step could be to implement the inception layer, implementing the forward and backward pass. Take a look at the existing layer to get an impression about how the inception layer could look like. Also, you can test the layer with the existing cnn model by using e.g. using the MNIST dataset. Afterwards, we should discuss how the inception model should look like, e.g. what's the best way
< zoq>
to branch out to add additional classifier.
ashutosh has joined #mlpack
ashutosh has quit [Client Quit]
< nilay>
zoq: ok thanks, i get the idea.
dms has quit [Ping timeout: 250 seconds]
nilay has quit [Ping timeout: 250 seconds]
new has joined #mlpack
< new>
hello friends
< new>
i'm new to this mlpack .... i built it
< new>
please tell me how to get started
< new>
i already have a decent knowledge on machine learning algorithms
< new>
hey there anybody?
< rcurtin>
new: I think three or four people have already suggested things that you could look at