svm-light的源代码

SVMlight by Thorsten Joachims http://www-ai.cs.uni-dortmund.de/SOFTWARE/SVM_LIGHT/svm_light.eng.html incorporating The PR_LOQO optimizer by Alex Smola http://spigot.anu.edu.au/~smola/ Ported to Windows by Martin Sewell http://www.martinsewell.com (" usage: svm_learn [options] example_file model_file\n\n"); ("Arguments:\n"); (" example_file-> file with training data\n"); (" model_file -> file to store learned decision rule in\n"); ("General options:\n"); (" -? -> this help\n"); (" -v [0..3] -> verbosity level (default 1)\n"); ("Learning options:\n"); (" -c float -> C: trade-off between training error\n"); (" and margin (default 1000)\n"); (" -j float -> Cost: cost-factor, by which training errors on\n"); (" positive examples outweight errors on negative\n"); (" examples (default 1)\n"); (" -b [0,1] -> use biased hyperplane (i.e. x*w+b>0) instead\n"); (" of unbiased hyperplane (i.e. x*w>0) (default 1)\n"); (" -i [0,1] -> remove inconsistent training examples\n"); (" and retrain (default 0)\n"); ("Transduction options:\n"); (" -p [0..1] -> fraction of unlabeled examples to be classified\n"); (" into the positive class (default is the ratio of\n"); (" positive and negative examples in the training data)\n"); ("Kernel options:\n"); (" -t int -> type of kernel function:\n"); (" 0: linear (default)\n"); (" 1: polynomial (s a*b+c)^d\n"); (" 2: radial basis function exp(-gamma ||a-b||^2)\n"); (" 3: sigmoid tanh(s a*b + c)\n"); (" 4: user defined kernel from kernel.h\n"); (" -d int -> parameter d in polynomial kernel\n"); (" -g float -> parameter gamma in rbf kernel\n"); (" -s float -> parameter s in sigmoid/poly kernel\n"); (" -r float -> parameter c in sigmoid/poly kernel\n"); (" -u string -> parameter of user defined kernel\n"); ("Optimization options:\n"); (" -q [2..%d] -> maximum size of QP-subproblems (default 10)\n",nx); (" -m [5..] -> size of cache for kernel evaluations in MB (default 40)\n"); (" The larger the faster...\n"); (" -e float -> eps: Allow that error for termination criterion\n"); (" [y [w*x+b] - 1] >= eps (default 0.001)\n"); (" -h [5..] -> number of iterations a variable needs to be\n"); (" optimal before considered for shrinking (default 100)\n"); (" -f [0,1] -> do final optimality check for variables removed\n"); (" by shrinking. Although this test is usually \n"); (" positive, there is no guarantee that the optimum\n"); (" was found if the test is omitted. (default 1)\n"); ("Output options:\n"); (" -l char -> file to write predicted labels of unlabeled\n"); (" examples into after transductive learning\n"); (" -a char -> write all alphas to this file after learning\n"); (" (in the same order as in the training set)\n\n");