密度峰值聚类算法(Density peaks clustering,DPC)是一种基于密度的新型聚类算法。该算法的优点十分显著:所需参数较少,没有迭代过程,能自适应获得聚类数目并识别任意形状的簇类。该算法也有一些问题亟待解决:(1)在决策图上人工选择聚...密度峰值聚类算法(Density peaks clustering,DPC)是一种基于密度的新型聚类算法。该算法的优点十分显著:所需参数较少,没有迭代过程,能自适应获得聚类数目并识别任意形状的簇类。该算法也有一些问题亟待解决:(1)在决策图上人工选择聚类中心,产生聚类误差;(2)在密度不同的流形数据集上聚类效果不佳。针对这些不足,该文提出一种基于电子分层模型和凝聚策略的密度峰值聚类算法(Density peaks clustering based on electronic shells model and merging strategy,EMDPC)。其利用电子分层模型计算每个数据点的局部密度,更易识别出低密度簇;通过子簇凝聚策略自适应识别簇类数目,降低了人工选取聚类中心时误差发生的概率;并且子簇凝聚策略能较好地解决DPC在密度不均匀的流形数据上聚类效果不佳的问题。实验分析表明基于电子分层模型和凝聚策略的密度峰值聚类算法具有较高的精度和较好的聚类性能,其结果优于其他先进的聚类算法。展开更多
The thin-film thermoelectric cooler(TEC)is a promising solid-state heat pump that can remove the high local heat flux of chips utilizing the Peltier effect.When an electric current pulse is applied to the thin-film TE...The thin-film thermoelectric cooler(TEC)is a promising solid-state heat pump that can remove the high local heat flux of chips utilizing the Peltier effect.When an electric current pulse is applied to the thin-film TEC,the TEC can achieve an instantaneous lower temperature compared to that created by a steady current.In this paper,we developed a novel strategy to reduce the peak temperature of the chip working under dynamic power,thus making the semiconductor chip operate reliably and efficiently.A three-dimensional numerical model was built to study the transient cooling performance of the thin-film TEC on chips.The effects of parameters,such as the current pulse,the heat flux,the thermoelement length,the number of thermoelements,and the contact resistance on the performance of the thin-film TEC,were investigated.The results showed that when a current pulse of 0.6 A was applied to the thin-film TEC before the peak power of the chip,the peak temperature of the chip was reduced by more than 10℃,making the thin-film thermoelectric cooler a promising technology for the temperature control of modern chips with high peak powers.展开更多
文摘密度峰值聚类算法(Density peaks clustering,DPC)是一种基于密度的新型聚类算法。该算法的优点十分显著:所需参数较少,没有迭代过程,能自适应获得聚类数目并识别任意形状的簇类。该算法也有一些问题亟待解决:(1)在决策图上人工选择聚类中心,产生聚类误差;(2)在密度不同的流形数据集上聚类效果不佳。针对这些不足,该文提出一种基于电子分层模型和凝聚策略的密度峰值聚类算法(Density peaks clustering based on electronic shells model and merging strategy,EMDPC)。其利用电子分层模型计算每个数据点的局部密度,更易识别出低密度簇;通过子簇凝聚策略自适应识别簇类数目,降低了人工选取聚类中心时误差发生的概率;并且子簇凝聚策略能较好地解决DPC在密度不均匀的流形数据上聚类效果不佳的问题。实验分析表明基于电子分层模型和凝聚策略的密度峰值聚类算法具有较高的精度和较好的聚类性能,其结果优于其他先进的聚类算法。
基金the National Natural Science Foundation of China(Grant No.51778511)Natural Science Foundation of Hubei Province(Grant No.2018CFA029)+3 种基金Key Research and Design Projects of Hubei Province(Grant No.2020BAB129)Key Project of ESI Discipline Development of Wuhan University of Technology(Grant No.2017001)Scientific Research Foundation of Wuhan University of Technology(Nos.40120237 and 40120551)the Fundamental Research Funds for the Central Universities(WUT:2021IVA037)。
文摘The thin-film thermoelectric cooler(TEC)is a promising solid-state heat pump that can remove the high local heat flux of chips utilizing the Peltier effect.When an electric current pulse is applied to the thin-film TEC,the TEC can achieve an instantaneous lower temperature compared to that created by a steady current.In this paper,we developed a novel strategy to reduce the peak temperature of the chip working under dynamic power,thus making the semiconductor chip operate reliably and efficiently.A three-dimensional numerical model was built to study the transient cooling performance of the thin-film TEC on chips.The effects of parameters,such as the current pulse,the heat flux,the thermoelement length,the number of thermoelements,and the contact resistance on the performance of the thin-film TEC,were investigated.The results showed that when a current pulse of 0.6 A was applied to the thin-film TEC before the peak power of the chip,the peak temperature of the chip was reduced by more than 10℃,making the thin-film thermoelectric cooler a promising technology for the temperature control of modern chips with high peak powers.
