Soft-Reward Based Reinforcement Learning by Spiking Neural Networks
In this paper, we propose algorithm based reinforcement learning for spiking neural networks. The algorithm simulates biological adaptability and uses the soft-reward from environment to modulate the synaptic weight, which combines spike-timing-dependent plasticity (STDP), winner-take-all mechanism. The algorithm is tested to classify a number of standard benchmark dataset. The obtained results show the effectiveness of the proposed algorithm.
Helen Zhang, Gang Shen and David Jin
W. Y. Shi "Soft-Reward Based Reinforcement Learning by Spiking Neural Networks", Advanced Materials Research, Vols. 219-220, pp. 770-773, 2011