How is threshold value set in PCA ? Alternatively how to increase efficiency of PCA?
We are implementing a face detection-recognition,gender recognition. We are using PCA for detection of human face in images. Please tell us how to decide a threshold value for the PCA algorithm. Also any alternatively methods , how to increase efficiency of PCA?
no silver bullet here, you will have to run tests on your own data.
the predict method returns a 'confidence' value, that's actually the distance to the closest item in your trainset. a possible threshold value for rejection will be somewhere between the largest true and the smallest false recognition result