论文部分内容阅读
本文提出一个新的解线性规划的Hopfields-型网络。该网络基于线性规划的对偶理论,并使用了Sigmoid函数,但不需要预先给定的罚参数和乘法模拟器。我们证明该网络不仅全局收敛到线性规划的精确解,而且能同时解原规划和对偶规划。由于在该网络中没有使用乘法模拟器而利用了Sigmoid函数,因此该模型是很容易用硬件实现的。
This paper presents a new Hopfields-type network for linear programming. The network is based on the duality theory of linear programming and uses the Sigmoid function, but does not require a given penalty parameter and multiplier simulator. We prove that the network not only globally converges to the exact solution of linear programming, but also can solve original planning and dual planning at the same time. The Sigmoid function is used because no multiplication simulator is used in the network, so the model is easily implemented in hardware.