A study about approximations to the Neyman–Pearson detector with SVMs is presented.
The functions approximated by C-SVMs and 2C-SVMs after training are obtained.
2C-SVMs can be used to approximate the optimum Neyman–Pearson detector.
PFA control is achieved varying the Ɣ parameter of 2C-SVM and prior probabilities.
Results with synthetic and real radar data are presented to validate the study.