A Novel Spiking Neural P System for Image Recognition
Xiantai Gou, Qifen Liu, Haina Rong, Meng Hu, Pirthwineel Paul, Fang Deng, Xihai Zhang and Zhibin Yu
Spiking neural P systems (SNPS), a kind of distributed parallel bioinspired model, has been a research hotspot in the field of membrane computing. SNPS has been widely concerned by scholars due to its powerful computing capacity and brain-like information transmission schema. Nevertheless, there are quite a few research results about SNPS with learning ability applied for image recognition. In this research, the routing mechanism in capsule neural network is introduced into SNPS to update the weights between synapses of spiking neurons dynamically. The learning ability of SNPS is realized by the weight update algorithm, which represents the changes in the strength of neuronal synaptic connections. Moreover, this is the first attempt to construct a novel universal network model of SNPS with learning ability which extracts features though the image convolution. The experimental results demonstrate that the recognition accuracy of the Mixed National Institute of Standards and Technology database, namely MNIST, reaches 95.87% and the recognition accuracy of English letters with noise and rotation reaches 98.06% in SNPS, which verify the feasibility and effectiveness of the model we constructed.
Keywords: Membrane computing, spiking neural P systems, image recognition, learning ability