In recent years, it has been demonstrated that spintronics devices can work as artificial synapses and neurons, which, when assembled in a topology inspired by the brain, considerably reduce the energy consumption of AI. However, artificial neural networks (ANNs), whether based on software or spintronics, retain considerable limitations. Most ANNs employed in current applications like image/sound recognition and data processing are trained specifically to perform a single task. This limitation is due to catastrophic forgetting, an exponential loss of memory upon learning a new task, while the memory loss in the brain is described by a power law. This issue is critical for human-machine interface applications where data is processed in real-time. Learning new tasks while remembering old ones is a trade-off between plasticity and rigidity: synaptic weights need to be modified to learn, but also to remain stable in order to remember.
Within METASPIN we will develop a new class of neuromorphic hardware that will use magneto-ionics to support synaptic metaplasticity, i.e. a feature inspired by the human brain based on assigning a ‘hidden value’ to the states of artificial synapses to encode how important each state is. This will make it easier or harder to reconfigure the synaptic state upon learning a new task, giving a hierarchy to previously learned information and thus preventing catastrophic forgetting. The synaptic states will be given by the two magnetisation orientations in ferromagnets with perpendicular magnetic anisotropy, and by ferro/antiferromagnetic order in materials where the two phases coexist. In all cases, magneto-ionic gating will be used to locally modulate intrinsic magnetic properties to assign ‘hidden states’ to each synaptic state. The magneto-ionic hidden states will translate into a modulation of the switching probability between synaptic states, introducing the metaplasticity functionality. In parallel, we will develop ANNs learning schemes, adapted to our device physics and inspired by biological synaptic activity, that can learn with mitigated catastrophic forgetting.