局域互联神经网络的关联存储

Locally interconnected neural network for associative memory

基于全局互联的Hopfield模型,本文提出了局域互联关联存储的新概念.与全局互联相比,局域互联具有较小的关联矩阵,因而,有利于用现有的空间光调制器加以实现.同时,计算机模拟结果表明,它仍然具有全局关联存储的能力.

Based on the globally interconnected Hopfield model, a new concept of locally interconnected associative memory is proposed in this paper. Compared with the globally interconnected, locally interconnected neural network has a smaller interconnection weight matrix (IWM), therefore, it can be easily implemented with commercially available spatial light modulators. Moreover, computer simulations indicate that it still possesses the global association ability.