A generalization of neural networks called Boolean networks isnconsidered. Random (n,k)-networks consist ofnn processors, each connected randomly to k others,ncomputing random k-input Boolean functions. The dynamicnbehavior of these networks has been studied extensively. The authorsnexamine the asynchronous dynamics of these networks, and prove thatnconvergence to fixpoints is assured for almost all random (n,nk)-networks at the limit n→∞, providednkn. The proof of this result lies in randomngraph theory
展开▼
机译:考虑了称为布尔网络的神经网络的一般化。随机( n e1>, k e1>)网络由n n e1>个处理器组成,每个处理器随机连接到其他 k e1>个处理器,n个计算随机 k e1>-输入布尔函数。这些网络的动态行为已被广泛研究。作者对这些网络的异步动力学进行了研究,并证明了几乎所有随机( n e1>,n k e1>)网络在极限 n < / e1>→∞,提供n k e1 >> log n e1>。该结果的证明在于随机图理论
展开▼