Abstract
Catastrophic forgetting (CF) refers to the sudden and severe loss of prior information in learning systems when acquiring new information. CF has been an Achilles heel of standard artificial neural networks (ANNs) when learning multiple tasks sequentially. The brain, by contrast, has solved this problem during evolution. Modellers now use a variety of strategies to overcome CF, many of which have parallels to cellular and circuit functions in the brain. One common strategy, based on metaplasticity phenomena, controls the future rate of change at key connections to help retain previously learned information. However, the metaplasticity properties so far used are only a subset of those existing in neurobiology. We propose that as models become more sophisticated, there could be value in drawing on a richer set of metaplasticity rules, especially when promoting continual learning in agents moving about the environment.
Catastrophic forgetting (CF) is the sudden and severe loss of memory for previously stored information due to the learning of new information. Standard artificial neural networks have traditionally suffered from CF when undertaking sequential learning tasks.A wide variety of strategies have been used to minimise forgetting, such as generative replay, constructive algorithms, complementary systems, sparse storage, and meta-learning.Metaplasticity is the activity-dependent plasticity of future plasticity or learning. Various network models use metaplasticity-type strategies to reduce forgetting during continual learning.Combining metaplasticity-type rules with other strategies often improves network performance even further.Models may benefit even more by drawing on other properties of biological metaplasticity phenomena.