如何实现原型的出生与死亡以进行OOD检测？

Out-of-Distribution (OOD) detection is essential for the secure deployment of machine learning models, with prototype-based learning methods being a mainstream strategy for achieving OOD detection. Existing prototype-based learning methods typically rely on a fixed number of prototypes, which fails to adapt to the inherent complexity differences across various categories. There is currently a lack of mechanisms that can adaptively adjust the number of prototypes based on data complexity. Inspired by the processes of cell birth and death in biology, a novel method named PID (Prototype bIrth and Death) is proposed to dynamically adjust the prototype count according to data complexity. This method relies on two dynamic mechanisms to enable more flexible prototype management.