It is for method evaluation. Do you know about FMR and FNMR and other evaluation parameters?
handbook of fingerprint recognition described them. the lower EER is better.
They are performance metrics. However, the ROC is generated by plotting the detection rate versus the false positive rate, which means ROC involves two performance metrics. If you are going to compare the performance of two detection algorithms, the ROC may be not a good metric. If the ROC of algorithm 1 is above the ROC of algorithm 2, algorithm 1 is better than algorithm 2. However, if the two ROC cross at some point, it is hard to say which one is better. In such context, you may have to adopt some unique performance metric, such as Bayesian average cost, total error, the mutual information between the system input and output.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling