Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Opencv3.0 LogisticRegression svm scaling

I am trying to use cv::ml::LogisticRegression to scale the weights given by a SVM. To train the model I am passing the weight data in the range of [-inf, inf] and the labels (-1 or 1) to the function cv::ml::LogisticRegression::train()

Till now everything works fine, but when I try to predict a sample by using cv::ml::LogisticRegression::predict() I allways get the result -1 or 1. Thus exactly the same result as from the SVM. In my opinion the result should be a probability between 0 and 1.

Used parameters for cv::ml::LogisticRegression():

  • learning rate: 0.001
  • iterations: 1000
  • train method: BATCH
  • regularisation: REG_L2
  • minibatch_size: 1

Am I doing anythin wrong ?