如何使用Libsvm来解决培训是否正确?

时间:2017-05-07 08:53:05

标签: matlab svm libsvm

我已经开始使用LibSVM训练两个类了。

训练数据的安排如下:
训练两个班级的数据(每个班级有7128个图像):( 14256 X 2304)
培训两个班级的标签(每个班级有7128张图片(1和-1)):( 14256 X 1)

model = svmtrain(training_label, training_data, '-s 0 -t 0 -c 1 -g 0.1')

我收到了这个输出:

..
WARNING: using -h 0 may be faster
*..*
optimization finished, #iter = 4625
nu = 0.024285
obj = -173.499920, rho = 13.995985
nSV = 577, nBSV = 189
Total nSV = 577


model = 

Parameters: [5x1 double]
  nr_class: 2
   totalSV: 577
       rho: 13.9960
     Label: [2x1 double]
sv_indices: [577x1 double]
     ProbA: []
     ProbB: []
       nSV: [2x1 double]
   sv_coef: [577x1 double]
       SVs: [577x2304 double]

特征向量:

figure
pos = find(training_label == 1);
neg = find(training_label == -1);
plot(training_data(pos,1), training_data(pos,2), 'ko', 'MarkerFaceColor', 'b'); hold on;
plot(training_data(neg,1), training_data(neg,2), 'ko', 'MarkerFaceColor', 'g')

enter image description here

我的问题是如何检查培训是否正确完成?什么是输出显示?

0 个答案:

没有答案