strin / bayespa Goto Github PK

Hi, is there any way of obtaining some description of the topics learned by this model (for example, a list of top words)?

when I enter
$sudo python setup.py install
I got this follow blew:

/usr/lib/python2.7/distutils/dist.py:267: UserWarning: Unknown distribution option: 'zip_safe'
warnings.warn(msg)
running install
running build
running build_py
creating build
creating build/lib.linux-x86_64-2.7
creating build/lib.linux-x86_64-2.7/medlda
copying medlda/init.py -> build/lib.linux-x86_64-2.7/medlda
running build_ext
building 'libbayespagibbs' extension
creating build/temp.linux-x86_64-2.7
creating build/temp.linux-x86_64-2.7/medlda
creating build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA
creating build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src
creating build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/ap.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/ap.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/cholesky.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/cholesky.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/cokus.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/cokus.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/InverseGaussian.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/InverseGaussian.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/Mapper.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Mapper.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/MVGaussian.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/MVGaussian.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/objcokus.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/objcokus.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/spdinverse.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/spdinverse.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/Document.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Document.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/utils/Sample.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Sample.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDA.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDA.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2 -Wall -Wstrict-prototypes -fPIC -DMAJOR_VERSION=0 -DMINOR_VERSION=1 -Imedlda/OnlineGibbsMedLDA/inc -Imedlda/OnlineGibbsMedLDA/inc/utils -I/usr/include/python2.7 -c medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDAWrapper.cpp -o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDAWrapper.o -std=c++11 -w
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++ [enabled by default]
g++ -pthread -shared -Wl,-O1 -Wl,-Bsymbolic-functions -Wl,-Bsymbolic-functions -Wl,-z,relro build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/ap.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/cholesky.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/cokus.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/InverseGaussian.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Mapper.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/MVGaussian.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/objcokus.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/spdinverse.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Document.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/utils/Sample.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDA.o build/temp.linux-x86_64-2.7/medlda/OnlineGibbsMedLDA/src/OnlineGibbsMedLDAWrapper.o -L/opt/local/lib/ -lboost_python-mt -o build/lib.linux-x86_64-2.7/libbayespagibbs.so
/usr/bin/ld: cannot find -lboost_python-mt
collect2: error: ld returned 1 exit status
error: command 'g++' failed with exit status 1

libboost-all-dev is already the newest version.
gcc-4.8 is already the newest version.
python-numpy is already the newest version.
Plz help.

• refractor global sample representation
• implement multi-task sampling procedure
• implement multi-task wrapper

"gcc > 4.8" should be "gcc >= 4.8"

with the following setting

import medlda
log = open("20ng_result.txt", "a")
batchsize = 512
label = 20
numword = 53975
num_sample = 100
for k in range(20, 120, 20):
        pamedlda = medlda.OnlineGibbsMedLDA(num_topic = k, labels = label, words = numword, alpha=1/k)
        pamedlda.train_with_gml('/home/wenbo/mfs/data/20ng/20ng_train.gml', batchsize)
        (pred, ind, acc) = pamedlda.infer_with_gml('/home/wenbo/mfs/data/20ng/20ng_test.gml', num_sample)
        log.write("topic: %d, batch: %d, numsample: %d, acc: %0.2f\n" % (k, batchsize, num_sample, acc))
log.close()

the results seem to differ from original paMedLDAgibbs implementation. Here the accuracy drops dramatically as the number of topics k increases.

topic: 10, batch: 512, numsample: 100, acc: 0.80
topic: 20, batch: 512, numsample: 100, acc: 0.80
topic: 30, batch: 512, numsample: 100, acc: 0.80
topic: 40, batch: 512, numsample: 100, acc: 0.80
topic: 50, batch: 512, numsample: 100, acc: 0.79
topic: 60, batch: 512, numsample: 100, acc: 0.78
topic: 70, batch: 512, numsample: 100, acc: 0.76
topic: 80, batch: 512, numsample: 100, acc: 0.72
topic: 90, batch: 512, numsample: 100, acc: 0.72
topic: 100, batch: 512, numsample: 100, acc: 0.70

Hi, thank you for this project! I was able to set up, train and infer this model on my own (binary classification) dataset.

However, infer_with_gml only provides me predicted labels; I would like to obtain a score that I can later use to construct an ROC curve for this model.

Could you guide me on how to modify the code to obtain scores instead of binary predictions?