libmf使用

主页介绍

https://www.csie.ntu.edu.tw/~cjlin/libmf/

github介绍

https://github.com/cjlin1/libmf

LIBMF is a library for large-scale sparse matrix factorization

providing solvers for real-valued matrix factorization, binary matrix factorization, and one-class matrix factorization

使用

下载zip file文件，也可以下载别的文件，解压后放到linux文件目录下

进入目录，输入“make”进行编译

real_matrix.tr.txt' 和 `real_matrix.te.txt'是训练和测试数据集

`binary_matrix.tr.txt' 和`binary_matrix.te.txt.'中<value>集是{-1, 1}

在libmf（mf-train所在目录下）目录下使用 ./mf-train以及./mf-predict：查看命令可以用的操作。训练命令。每次迭代

./mf-train demo/real_matrix.tr.txt model

• ./mf-train -l1 0.05 -l2 0.01 demo/real_matrix.tr.txt model
  ./mf-train -l1 0.015,0 -l2 0.01,0.005 demo/real_matrix.tr.txt model
  ./mf-train -f 5 -l1 0,0.02 -k 100 -t 30 -r 0.02 -s 4 demo/binary_matrix.tr.txt model
  ./mf-train -p demo/real_matrix.te.txt demo/real_matrix.tr.txt model #holdout
  ./mf-train -v 5 demo/real_matrix.tr.txt   #five fold cross validation
  ./mf-train -f 2 --nmf demo/real_matrix.tr.txt #do non-negative matrix factorization with generalized KL-divergenc

   

./mf-predict

./mf-predict -e 1 demo/real_matrix.te.txt model output

sh demo.sh可以运行demo