Synaptic metaplasticity in binarized neural networks

Unlike the brain, artificial neural networks, including state-of-the-art deep networks for computer vision, are subject to "catastrophic forgetting": they rapidly forget previous task when trained on a new one. Neuroscience suggests that biological synapses avoid this issue through process of synaptic consolidation and metaplasticity: plasticity itself changes upon repeated events. In work, we show concept metaplasticity can be transferred particular type binarized reduce catastrophic forgetting.