针对脱机文字识别,提出了一种基于线性合成的双粒度递归神经网络(Recurrent neural net work,RNN)集成系统.首先,使用单词RNN对未知图像进行识别;然后,依据识别结果进行字符分割,使用字符RNN对分割后的字符进行识别,并利用查表法计算字...针对脱机文字识别,提出了一种基于线性合成的双粒度递归神经网络(Recurrent neural net work,RNN)集成系统.首先,使用单词RNN对未知图像进行识别;然后,依据识别结果进行字符分割,使用字符RNN对分割后的字符进行识别,并利用查表法计算字符的后验概率;最后,综合两个RNN的识别结果决定最终单词输出.在CAPTCHA识别和手写识别上的实验结果证明了该系统的有效性.展开更多
With the development of Internet, traditional network is converting to active network computing, which makes the problem of mobile-code security emerge as one of the most challenges of research in Computer Science tod...With the development of Internet, traditional network is converting to active network computing, which makes the problem of mobile-code security emerge as one of the most challenges of research in Computer Science today. The literature introduces one method that solves the problem of mobile-code security: Language-Based Security,and decribes in detail several methods in this research area, such as Proof-Carrying Code(PCC) and Typed Assembly Language(TAL). Our resolvation of mobile-code safety is presented at last.展开更多
文摘针对异常丢包导致DTN(Delay Tolerant Networks)消息传递成功率降低、时延和开销比增大的问题,提出一种基于可信度的改进DTN安全路由算法。该算法通过收集节点的历史相遇记录,计算并依据节点可信度,在路由转发决策阶段优化转发节点的选择,从而提高消息传递成功率,减少网络资源的浪费,增强DTN抵御异常丢包的能力。仿真实验结果表明改进的算法能够有效抵御异常丢包,综合性能优于典型的DTN路由算法Spray and Wait和Prophet。
文摘针对脱机文字识别,提出了一种基于线性合成的双粒度递归神经网络(Recurrent neural net work,RNN)集成系统.首先,使用单词RNN对未知图像进行识别;然后,依据识别结果进行字符分割,使用字符RNN对分割后的字符进行识别,并利用查表法计算字符的后验概率;最后,综合两个RNN的识别结果决定最终单词输出.在CAPTCHA识别和手写识别上的实验结果证明了该系统的有效性.
文摘With the development of Internet, traditional network is converting to active network computing, which makes the problem of mobile-code security emerge as one of the most challenges of research in Computer Science today. The literature introduces one method that solves the problem of mobile-code security: Language-Based Security,and decribes in detail several methods in this research area, such as Proof-Carrying Code(PCC) and Typed Assembly Language(TAL). Our resolvation of mobile-code safety is presented at last.
