Self-backpropagation of synaptic modifications elevates the efficiency of spiking and artificial neural networks
SBP (Self-backpropagation)： The phenomenon of SBP represents a form of nonlocal activity–dependent synaptic plasticity that may endow developing neural circuits the capacity to modify the weights of input synapses on a neuron in accordance with the status of its output synapses. Inspired by this, we have made progress in local learning with self-backpropagation at the mesoscopic scale, where SBP can play a role in both SNN and ANN with low energy consumption.
T. Zhang, X. Cheng, S. Jia, M. Poo, Y. Zeng, B. Xu*, Self-backpropagation of Synaptic Modifications Elevates the Efficiency of Spiking and Artificial Neural Networks, Science Advances, 2021.
A brain-inspired algorithm that mitigates catastrophic forgetting of artificial and spiking neural networks with low computational cost.
NACA（Neuromodulation-Assisted Credit Assignment）： Borrowing from the role of neuromodulator levels in the brain in modifying synapses, we establish synaptic and neuronal selection pathways to modulate local synaptic plasticity. This enables low computational cost online learning capabilities in both SNNs and ANNs.
T. Zhang*, X. Cheng, S. Jia, C. T. Li, M. Poo, B. Xu*, A brain-inspired algorithm that mitigates catastrophic forgetting of artificial and spiking neural networks with low computational cost, Science Advances, 2023
