基于SCSO-GRU模型的网络流量预测

网络流量有实时性、不稳定性和时序相关性等特点,传统网络流量预测模型存在泛化能力不强和预测精度低等不足之处。为解决这些不足,本文提出一种结合基于正余弦的群优化(SCSO)算法的GRU神经网络的网络流量预测模型(SCSO-GRU)。首先,介绍SCSO算法的粒子更新原理;然后构建SCSO-GRU神经网络的网络流量预测模型,将SCSO算法用于模型训练,提高训练效果,克服传统GRU神经网络收敛于局部最优的缺点;最后用SCSO-GRU模型进行网络流量预测。实验结果表明,与传统LSTM和GRU模型相比,本文模型具有显著的收敛效果和较好的预测精度,可以更好地刻画网络流量变化趋势。

基于e-learning的医学科研管理模式在学术交流领域的实践与经验

在线学术交流方式如e-learning等,克服传统学术交流模式中学科领域、经费、空间和时间的限制,能够有效提升学术交流效率。当下,越来越多的科技工作者利用e-learning等在线学术交流平台,分享科技信息、交流学术观点、追踪最新科学研究趋势,形成高效学术交流模式。本文对医学科研管理人员运用e-learning等方式协助科技工作者进行高效、共享的学术交流,总结目前的实践应用现状及经验,并为未来e-learning在学术交流领域的扩大使用提供建设性启示。

Algorithm Design and Application of Laminar Cooling Feedback Control in Hot Strip Mill

Feedback control is one of the most important ways to improve coiling temperature control precision during laminar cooling process.Laminar cooling equipments of a hot strip mill and structure of the control system were introduced.Feedback control algorithm based on PI controller and that based on Smith predictor were designed and tested in a hot strip mill respectively.Practical application shows that the feedback control system based on PI controller plays a limited role in improving coiling temperature control precision.The feedback control system based on Smith predictor runs stable and reliable.When the measured coiling temperature deviates from the target value,it can be adjusted to the required range quickly and steadily by Smith predictor feedback control,which improves the coiling temperature control precision greatly,and qualities of hot rolled strips are improved significantly