还指出由于当前的连续小波神经网络主要使用传统BP神经网络的随机初始化方法和基于梯度的训练算法，因此存在收敛性差的缺点。
It is also indicated that current WNN has a poor convergence performance because of adopting the random initialization method and gradient training algorithm of traditional BP NET.
利用A5/1算法的密钥和帧序列号初始化弱点，可以用相关分析技术区分出A5/1密钥流与真随机比特序列或恢复A5/1密钥。
The A5/1 key-streams can be easily distinguished from the random sequences, or even the A5/1 keys can be recovered, because of the known weakness of the A5/1 algorithm.
应用推荐