Generalized Few-shot Learning

A deep neural network that learns novel classes from few samples per class, preventing catastrophic forgetting of base classes and ensuring classifier calibration.