我试图理解train.model.b中的sv_coef输出。我已经使用以下参数的libsvm库运行了svm train-
"libsvm-3.22/svm-train -g {0} -c 50 -t 2 -b 1 -h 0 c.train.scaled train.model.b"
在我的模型输出中,sv_coef要么是50,要么是任何非50的双值。
示例输出-
svm_type c_svc
kernel_type rbf
gamma 0.001
nr_class 2
total_sv 39164
rho 0.486988
label 1 -1
probA -1.55179
probB 0.726919
nr_sv 19156 20008
SV
50 1:-0.026824774 2:-0.12938578 3:-0.013288319 4:-0.49732185 5:-0.16337877 6:-0.042702758 7:-0.050470105 8:-0.39069185 9:-0.37822306 10:-0.036900605 11:-0.084576928 12:-0.030365458 13:-0.22727974 14:-0.072242554 15:-0.11087256 16:-0.082843699 17:-0.12381641 18:-0.096841586 19:-0.13324288 20:-0.019173553 21:-0.024287144 22:-0.091258826 23:-0.083265353..............
19.74060492637799 1:-0.026824774 2:-0.12938578 3:-0.013288319 4:-0.49732185 5:-0.16337877 6:-0.042702758 7:-0.050470105 8:-0.39069185 9:-0.37822306 10:-0.036900605 11:-0.084576928 12:-0.030365458 13:-0.22727974 14:-0.072242554 15:-0.11087256 16:-0.082843699 17:-0.12381641 18:-0.096841586 19:-0.13324288 20:-0.019173553 21:-0.024287144 22:-0.091258826 23:-0.083265353......................
我需要理解,为什么sv系数是50或任何非50的双值?
