多值光谱自联想反馈神经网络

Multivalued Spectral Associative Autonomous Recurrent Network

将神经元状态取值的多样化和光谱自联想反馈神经网络的理论相结合,提出了多值光谱自联想反馈神经网络,有效扩展了Ronald G.Spencer提出的光谱自联想反馈神经网络.多值光谱自联想反馈神经网络的优势,主要在于有效扩大了网络神经元状态的取值范围,使其不仅能存储两级样本,亦可存储一些其他值,这样能使神经元之间的联系更为紧密.实验结果亦表明其优越性更加适应于在通信领域中应用.

In this paper, the multivalued spectral associative autonomous recurrent network is presented, which is a extension of the spectral associative autonomous recurrent neural network, SARN in Ronald G. Spencer. Compared with SARN, MSARN has two main advantages, wider value extension and loser interaction. The experimental results manifest these advantages.