RETRIEVAL PROPERTIES OF HOPFIELD NEURAL NETWORK MODELS
MA YU-QIANG, ZHANG YUE-MING, GONG CHANG-DE, (1)南京大学物理系,南京210008; (2)南京医学院物理系,南京210029; (3)中国高等科学技术中心,北京100080 · Acta Physica Sinica · 1993
In this paper, we propose a bimodal distribution of random neuronal activity threshold with different probabilities, to consider the influences on the retrieval properties of neural network. It is shown that the system successfully retrieves information even if the number of stored patterns exceeds the critical value of the pure Hopfield model.