【摘要】 提出了一种基于改进型树型奇偶机的密钥交换方法。利用伪随机数产生器为通信双方产生共同的输入向量，降低网络同步所需的数据量；使用基于滑动窗口的学习规则来动态改变学习率，降低同步次数；并且进一步提出基于背包算法和深度优先搜索遍历算法的高效输入序列生成方法，增大正向学习概率加快同步速度。仿真实验结果表明，改进模型的通信数据量下降了60%以上，同步时间显著下降。更多还原

【Abstract】 This paper proposes a key exchange method based on improved tree parity machine. The pseudo-random number generator is used to generate common input vectors for both sides of communication,so as to reduce the amount of data needed in network synchronization. Sliding window based learning rules are used to dynamically change the learning rate,so as to reduce the number of synchronization. Furthermore,an efficient input sequence generation method based on knapsack algorithm and depth first search traversal algorithm is proposed to increase the positive learning probability and accelerate the synchronization speed. The simulation results show that the communication data volume of the improved model is reduced by more than 60%,and the synchronization time is significantly reduced.更多还原