This repo is forked and refactored from the popular libsvm https://github.com/cjlin1/libsvm for CSE 569S.

Changes made in the framework include:

• Removed interfaces to other languages.
• Removed data expansion function to simplify the program.
• Added a sample dataset from the original program's website.
• Added a cmake file to simplify building process.
• Replaced deprecated libraries.
• Reformatted code.
• And more.

If you don't like this simplified version, you are welcome to directly use the original program from the link above.

Generate Makefile

cmake . 

Build the program

make

Please see the original repo and its website https://www.csie.ntu.edu.tw/~cjlin/libsvm/ for detailed instructions.

./svm-train data/a1a test_model

optimization finished, #iter = 537 nu = 0.460270 obj = -673.031415, rho = 0.628337 nSV = 754, nBSV = 722 Total nSV = 754

./svm-predict data/a1a.t test_model output

Accuracy = 83.5864% (25875/30956) (classification)