Support Vector Machine (SVM)

From: 台大林智仁 教授   

• guide for beginners.(A Practical Guide to Support Vector Classification)

• SVM slides
• LIBSVM -- A Library for Support Vector Machines
• Other Documents of LIBSVM by Users

• Download libsvm

  • libsvm.zip(For MS-Window)

    • svmtoy.exe
      • D:\libsvm-2.84\windows>svmtoy.exe

  • libsvm.tar.gz(For Linux/Unix)

• Frequently Asked Questions (FAQ)
  • The prediction rate is low. How could I improve it?
  • Practice
    • heart
      • D:\libsvm-2.84\windows>svmtrain.exe heart.txt
      • D:\libsvm-2.84\windows>svmpredict.exe heart.txt heart.txt.model heart.txt.output
    • heart_scale
      • D:\libsvm-2.84\windows>svmscale.exe heart.txt > heart.scale
      • D:\libsvm-2.84\windows>svmtrain.exe heart.scale
      • D:\libsvm-2.84\windows>svmpredict.exe heart.scale heart.scale.model heart.scale.output
  • Practice
    • D:\libsvm-2.84\windows>svmtrain glass.scale
    • D:\libsvm-2.84\windows>svmpredict glass.scale  glass.scale.model  glass.scale.output
      Accuracy = 60.2804% (129/214) (classification)
  • Practice
    • LIBSVM data sets
  • an easy script (easy.py) for users who know NOTHING about svm. It makes everything automatic--from data scaling to parameter selection.
    The parameter selection tool grid.py generates the following contour of cross-validation accuracy.
    • To use this tool, you also need to install python (Python 2.5.1 (April 18, 2007)) and gnuplot.

SVM, Support Vector Machine