Furthermore, we find this mechanism to naturally prevent catastrophic forgetting, which as we show, is still an issue for the Hebbian/anti-Hebbian network

We propose a biologically plausible mechanism which updates neurons asynchronously and empirically find that this allows plasticity applied in pair with the dynamics to approximate the objective of classical Hebbian/anti-Hebbian networks

We find that in Hebbian/anti-Hebbian networks the effect of continuously applying plasticity differs from applying plasticity updates only when the network has reached its equilibrium state. The latter has been shown to enable representation learning (e.g. work by Pehlevan, Krotov and others).