PlaStIL: Plastic and Stable Memory-Free Class-Incremental Learning. (arXiv:2209.06606v1 [cs.CV])

Plasticity and stability are needed in class-incremental learning in order to
learn from new data while preserving past knowledge. Due to catastrophic
forgetting, finding a compromise between these two properties is particularly
challenging when no memory buffer is available. Mainstream methods need to
store two deep models since they integrate new classes using fine tuning with
knowledge distillation from the previous incremental state. We propose a method
which has similar number of parameters but distributes them differently in
order to …

arxiv free incremental memory plastic