mRMR_0.9_compiled

所属分类:matlab编程
开发工具:matlab
文件大小:481KB
下载次数:129
上传日期:2009-02-05 10:21:24
上 传 者reallycsc
说明:  mRMR(min-redundancy max-relevance)的matlab程序
(matlab program of mRMR(min-redundancy max-relevance))

文件列表:
mRMR_0.9_compiled\mi_0.9\condentropy.m (1034, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\condmutualinfo.m (1237, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\demo_mi.m (724, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\entropy.m (761, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estcondentropy.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estcondentropy.mexglx (125024, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estcondentropy.mexmac (54608, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\estentropy.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estentropy.mexglx (7255, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estentropy.mexmac (53908, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\estjointentropy.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estjointentropy.mexglx (124769, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estjointentropy.mexmac (54608, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\estmutualinfo.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estmutualinfo.mexglx (125610, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estmutualinfo.mexmac (54644, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\estpa.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estpa.mexglx (127488, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estpa.mexmac (59216, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\estpab.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estpab.mexglx (128216, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\estpab.mexmac (59180, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\findjointstateab.dll (40960, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\findjointstateab.mexglx (129114, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\findjointstateab.mexmac (59260, 2007-03-23)
mRMR_0.9_compiled\mi_0.9\jointentropy.m (924, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\makeosmex.m (599, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\mergemultivariables.m (1099, 2007-01-29)
mRMR_0.9_compiled\mi_0.9\mutualinfo.m (604, 2007-01-29)
mRMR_0.9_compiled\mrmr_mibase_d.m (363, 2007-01-29)
mRMR_0.9_compiled\mrmr_mid_d.m (1277, 2007-01-29)
mRMR_0.9_compiled\mrmr_miq_d.m (1351, 2007-01-29)
mRMR_0.9_compiled\mi_0.9 (0, 2008-11-26)
mRMR_0.9_compiled (0, 2008-11-26)

********************************************************** To run the mRMR variable/feature selection for discrete variables, you need to add the path of mi_0.9 in your Matlab path. Then simply run mrmr_mid_d.m or mrmr_miq_d.m for your data. If you have continuous data as input, you need to discretize the data first. Simple methods to do this can be thresholding at the mean or mean-plus/minus-std. Of course, you may also try the MI computation for continuous variables directly, as I showed in the my paper (below). However, typically the results are not as good as discrete variables. Note that this version is old and uses double-precision in mutual information computation, thus the feature selection results may be slightly different if you also compare the against those produced by the newer C versions downloadable from our website. The C versions uses single precision for floating numbers to save some memories. The codes cannot be re-distributed without permission from the author, Hanchuan Peng. We hope you cite our work as follows, which you can download the paper at Hanchuan Peng's web site http://research.janelia.org/peng (you may google and find out the latest website). Hanchuan Peng, Fuhui Long, and Chris Ding, "Feature selection based on mutual information: criteria of max-dependency, max-relevance, and min-redundancy," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 27, No. 8, pp.1226-1238, 2005. Should you have any question, please send email to hanchuan.peng@gmail.com or pengh@janelia.hhmi.org . **********************************************************

近期下载者

相关文件


收藏者