AN ONLINE SUPERVISED LEARNING ALGORITHM BASED ON FEEDBACK ALIGNMENT FOR MULTILAYER SPIKING NEURAL NETWORKS

The feedback alignment provides a biologically plausible learning mechanism, which can directly transmit error signals with a random weight matrix to multiple layers of a neural network. This paper proposes an online supervised learning algorithm based on the feedback alignment mechanism for multilayer spiking neural networks, named Multi-OSLFA, which can support real-time learning for the spatio-temporal pattern of spike trains. The online learning rule is represented by the kernel function of spike trains and adjusts the synaptic weights when the output neuron fires a spike during the miming process of spiking neural networks. The Multi-OSLFA algorithm is successfully applied to spike train learning tasks and nonlinear pattern classification problems on two UCI datasets. Simulation results indicate that the proposed algorithm can improve learning accuracy in comparison with other supervised learning algorithms. It shows that the proposed learning algorithm is effective for solving spatio-temporal pattern learning problems.