This model demonstrates how catastrophic forgetting in continual reinforcement learning can be prevents by using to synaptic weight components. In addition, experimental results from tagging, cross-tagging and roving experiments are reproduced.
Literature: P. Leimer, M. Herzog, W. Senn (April 2019). Synaptic weight decay with selective consolidation enables fast learning without catastrophic forgetting, bioRxiv, https://doi.org/10.1101/613265