A study on the plasticity of neural networks
This addresses a potential bottleneck in continual learning by highlighting how fine-tuning can negatively affect model performance, which is incremental as it builds on recent observations.
The paper investigates the loss of plasticity in neural networks during fine-tuning, where pretrained models may not achieve the same generalization as freshly initialized ones, and explores its implications for continual learning.
One aim shared by multiple settings, such as continual learning or transfer learning, is to leverage previously acquired knowledge to converge faster on the current task. Usually this is done through fine-tuning, where an implicit assumption is that the network maintains its plasticity, meaning that the performance it can reach on any given task is not affected negatively by previously seen tasks. It has been observed recently that a pretrained model on data from the same distribution as the one it is fine-tuned on might not reach the same generalisation as a freshly initialised one. We build and extend this observation, providing a hypothesis for the mechanics behind it. We discuss the implication of losing plasticity for continual learning which heavily relies on optimising pretrained models.