**The code requires prior installation of libinear
The windows 64 bit binary of 'train' is already included in the sleec_train folder
To run on other platforms, compile liblinear and use the matlab train executable for linux in sleec_train
For parallelization, we have used openmp library, just make sure that the compiler you are using
is OpenMP compatible**
To compile SLEEC, run the following command on the Matlab prompt
> make_SLEEC
SLEEC training and prediction code have been combined with the SLEEC functions. To run SLEEC, try
> load data.mat
> SLEECparams
> [result] = SLEEC(data, params);
The following explains the matlab structure for the data, parameters and the output result
SLEEC Data
data : structure containing the complete dataset
data.X : sparse n x d matrix containing train features
data.Xt : sparse m x d matrix containing test features
data.Y : sparse n x L matrix containing train labels
data.Yt : sparse m x L matrix containing test labels
SLEEC Parameters
params.num_learners : number of SLEEC learners in ensemble (default 10)
params.num_clusters : Initial number of clusters (default 300)
params.num_threads : Number of threads for parallelization (default 32)
params.SVP_neigh : Number of nearest neighbours to be preserved (default 15)
params.out_Dim : embedding dimensions (default 50)
params.w_thres : 1-w_thresh is the sparsity of regressors w (default 0.7)
params.sp_thresh : 1-sp_thresh is the sparsity of embeddings (default 0.7)
params.cost : liblinear cost coefficient (default 0.1)
params.NNtest : number of nearest neighbours to consider while testing (default 10)
params.normalize : 1 for normalized data, 2 for unnormalized data (only for mediamill)
params.fname : filename for logging purposes
SLEEC Result
result.clusterCenters : cluster centers for the different learners
result.tim_clus : time taken for clustering
result.SVPModel : model for different learners containing embeddings and regressors
result.SVPtime_mat : time taken for performing SVP for each learner
result.regressiontime_mat : time taken for learning regressors
result.precision : overall precision accuracy
result.predictAcc : precision accuracy per test point
result.predictLabels : Top-k labels predicted per test point
result.tim_test : time taken for the testing procedure
result.test_KNN : kNN Matrix for the test points
Toy Example
The zip file containing the source code also include the BibTeX dataset as a toy example. The following are the instructions to run SLEEC on the BibTeX dataset
> cd Toy_Example
> load bibtex.mat
> bibtexParams
> cd ..
> [result] = SLEEC(data, params);
To verify that you have run the code successfully, please compare the precision that you obtain with the result structure that has been provided in the Toy_Example folder.
没有合适的资源?快使用搜索试试~ 我知道了~
SLEECcode_多标签分类_源码
共188个文件
m:57个
mexw64:20个
cpp:19个
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 153 浏览量
2021-10-02
02:31:20
上传
评论
收藏 9MB ZIP 举报
温馨提示
Sparse Local Embeddings for Extreme Multi-label Classification——SLEEC源代码
资源详情
资源评论
资源推荐
收起资源包目录
SLEECcode_多标签分类_源码 (188个子文件)
XonOmega.c 26KB
XonOmegaTranspose.c 23KB
updateSparse.c 5KB
reorth_mex.c 5KB
bdsqr_mex.c 5KB
reorth.c 4KB
smvp.c 4KB
compOmegaYx.c 1KB
compOmegaYtx.c 1KB
dbdqr.c 1KB
tqlb_mex.c 1KB
compute_X_Omega.c 1KB
compute_X_Omega.c 1008B
compute_USV_Omega.c 1002B
compute_UdSV_Omega.c 972B
compute_USV_Omega 953B
smat.cpp 10KB
smat.cpp 10KB
smat.cpp 10KB
smat.cpp 10KB
findKNN_test.cpp 6KB
kmeansDP_FtSp.cpp 5KB
findKNN_rf_ed.cpp 5KB
smat_write.cpp 5KB
smat_write.cpp 5KB
smat_write.cpp 5KB
smat_write.cpp 5KB
evalPrec_rf.cpp 5KB
findKNN_rf_dp.cpp 5KB
identifyClusterDP_FtSp_sparse.cpp 4KB
formScoreMat.cpp 4KB
read_data.cpp 4KB
updateU.cpp 2KB
updateV.cpp 2KB
write_data.cpp 1KB
lanbpro.doc 3KB
lanpro.doc 3KB
laneig.doc 2KB
lansvd.doc 2KB
LICENSE.docx 33KB
bdsqr.exp 42B
tqlb.f 5KB
reorth.f 3KB
dbdqr.F 445B
smat.h 4KB
smat.h 4KB
smat.h 4KB
smat.h 4KB
bdsqr.lib 3KB
lanbpro.m 23KB
lanpro.m 14KB
lansvd.m 12KB
laneig.m 9KB
test.m 9KB
mmread.m 7KB
mmwrite.m 6KB
install_mex.m 4KB
clusterSVP_asym.m 3KB
mminfo.m 3KB
reorth_complex.m 3KB
test_PROPACK.m 3KB
reorth.m 3KB
reorth_m.m 3KB
update_gbound.m 3KB
test_MEX.m 3KB
SLEEC.m 2KB
findBestMultiply.m 2KB
XonOmega.m 2KB
multiplePrediction_lin.m 2KB
testtqlb.m 2KB
compute_int.m 1KB
updateSparse_slow.m 1KB
WAltMin_asymm.m 1KB
predict_learners.m 1KB
multipleSVP_lin.m 1KB
multipleClustering.m 1KB
hierKmeansFt.m 1KB
make.m 1KB
bdsqr.m 986B
refinebounds.m 939B
evalModelSize.m 917B
evalModelSize.m 917B
tqlb.m 852B
evalModelSize_bak.m 840B
runHierKmeansFt.m 781B
computeW.m 621B
pythag.m 618B
test_clusAssignFt_sparse.m 616B
evalnDCG.m 614B
thresh.m 469B
runKmeansDPFt.m 455B
evalPrecision.m 369B
getScoreMat.m 357B
make.m 337B
Cfunc.m 293B
readData.m 283B
SLEECparams.m 267B
normalizeMatrix.m 254B
normalizeMatrix.m 254B
normalizeMatrix.m 254B
共 188 条
- 1
- 2
摇滚死兔子
- 粉丝: 54
- 资源: 4227
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0