Das Gleich Fehlerquote (EER) is a critical performance metric in the field of biometric systems and maschinellem Lernen, particularly in classification tasks. It represents the point at which the Falsch-Akzeptanz-Rate (FAR) and the Falsch-Ablehnungs-Rate (FRR) are equal. In simpler terms, EER helps to find a balance between accepting genuine users and rejecting impostors.
When a biometric system is deployed, it must be able to accurately differentiate between valid and invalid users. The Falsch-Akzeptanz-Rate (FAR) is the rate at which unauthorized users are incorrectly accepted as legitimate, while the Falsch-Ablehnungs-Rate (FRR) is the rate at which legitimate users are incorrectly denied access. The EER provides a single value that indicates how well the system performs overall by showing where these two error rates meet.
In practical terms, a lower EER indicates a more accurate biometric system, as it signifies fewer errors in both acceptance and rejection. The EER is particularly useful for comparing different biometric systems or algorithms, as it provides a standardized measure of performance regardless of the specific application or context.
To visualize the EER, it is often plotted on a receiver operating characteristic (ROC) curve, where the x-axis represents the False Positive Rate, and the y-axis represents the true positive rate. The point at which the FAR and FRR intersect on this graph denotes the EER.