基于注意力机制与文本信息的用户关系抽取

随着社交媒体的发展,用户之间的关系网络对于社交媒体的分析有很大的帮助。因此,该文主要研究用户好友关系检测。以往的关于用户好友关系抽取的研究主要基于社交媒体上的结构化信息,比如其他好友关系,用户的不同属性等。但是,很多时候用户本身并没有大量的好友信息存在,同时也不一定有很多确定的属性。因此,我们希望基于用户发表的文本信息来对用户关系进行预测。不同于以往的潜在好友推荐算法,该文提出了一种基于注意力机制以及长短时记忆网络(long short-term memory,LSTM)的好友关系预测模型,将好友之间的评论分开处理,通过分析用户之间的评论来判断是否具备一定的好友关系。该模型将好友双方信息拼接后的结果作为输入,并将注意力机制应用于LSTM的输出。实验表明,用户之间的评论对于好友关系预测确实有较大的实际意义,该文提出的模型较之于多个基准系统的效果,取得了明显的提升。在不加入任何其它非文本特征的情况下,实验结果的准确率达到了77%。     

大系统分散鲁棒容错控制的进展及主要成果

综述了近２０年来关联大系统分散鲁棒容错控制的研究动态及取得的主要成果。说明了统一地研究分散鲁棒容错控制的必要性。在分散控制、鲁棒控制和容错控制研究领域内，分析了各种设计方法的主导思想及实质；指出了这些研究成果的意义与存在的问题。文中着重论述了分散、鲁棒、容错控制研究方法的发展途径。     

Modeling and CW operation of a quasi-three-level 946 nm Nd: YAG laser

A model is developed for an end-pumped quasi-three-level laser with population in the lower laser level at equilibrium such as for transitions to the manifolds⁴I9/2in Nd³⁺,⁴I15/2in Er³⁺,⁵I8in Ho³⁺and³H6in Tm³⁺. It is shown that the effect of residual lower laser level population on laser operation can be treated as a saturable loss. Room temperature operation on the⁴F3/2-⁴I9/2transition in Nd:YAG under CW dye laser pumping has been demonstrated with a threshold as low as 11.5 mW incident power and a slope efficiency of 7 percent with 0.3 percent output coupling. Performance is limited by the low output coupling and diffraction loss.     

A dual-mode multiuser detector for DS-CDMA systems

We introduce a dual-mode multiuser detector that dynamically switches its detection mode between matched-filter and decorrelator operations based on the channel characteristics. This detector significantly reduces the overall computational requirement while maintaining similar performance as that of the decorrelator. The switching mechanism of our dual-mode detector is designed by exploiting the performance-complexity tradeoff between the decorrelator and the matched-filter. Extensions of this idea to other types of multiuser detectors are also proposed     

Next-day electricity-price forecasting using a hybrid network

The paper proposes a novel method of forecasting short-term electricity price based on a two-stage hybrid network of self-organised map (SOM) and support-vector machine (SVM). In the first stage, a SOM network is applied to cluster the input-data set into several subsets in an unsupervised manner. Then, a group of SVMs is used to fit the training data of each subset in the second stage in a supervised way. With the trained network, one can predict straightforwardly the next-day hourly electricity prices. To confirm its effectiveness, the proposed model has been trained and tested on the data of historical energy prices from the New England electricity market.     

Ring‐opening polymerization of D,L‐lactide by rare earth 2,6‐dimethylaryloxide

Ring‐opening polymerization of D,L ‐lactide (LA) has been successfully carried out by using rare earth 2,6‐dimethylaryloxide (Ln(ODMP)₃) as single component catalyst or initiator for the first time. The effects of different rare earth elements, solvents, monomers and catalyst concentration as well as polymerization temperature and time on the polymerization were investigated. The results show that La(ODMP)₃ exhibits higher activity to prepare poly(D,L ‐lactide) (PLA) with a viscosity molecular weight of 4.5 × 10⁴ g mol^?1 and the conversion of 97 % at 100 °C in 45 min. The catalytic activity of Ln(ODMP)₃ has following sequence: La > Nd > Sm > Gd > Er > Y. A kinetic study has indicated that the polymerization is first order with respect to both monomer and catalyst concentration. The apparent activation energy of the polymerization of LA with La(ODMP)₃ is 69.6 kJ mol^?1. The analyses of polymer ends indicate that the LA polymerization proceeds according to ‘coordination–insertion’ mechanism with selective cleavage of the acyl–oxygen bond of the monomer. Copyright © 2004 Society of Chemical Industry     

Effect of surfactants and liquid hydrocarbons on gas hydrate formation rate and storage capacity

Hydrate formation rate plays an important role in making hydrates for the storage and transport of natural gas. Micellar surfactant solutions were found to increase gas hydrate formation rate and storage capacity. With the presence of surfactant, hydrate could form quickly in a quiescent system and the energy costs of hydrate formation reduced. Surfactants (an anionic surfactant, a non‐ionic surfactant and their mixtures) and liquid hydrocarbons (cyclopentane and methylcyclohexane) were used to improve hydrate formation. The experiments of hydrate formation were carried out in the pressure range 3.69–6.82 MPa and the temperature range 274.05–277.55 K. The experimental pressures were kept constant during hydrate formation in each experimental run. The effect of anionic surfactant (sodium dodecyl sulphate (SDS)) on natural gas storage in hydrates is more pronounced compared to a non‐ionic surfactant (dodecyl polysaccharide glycoside (DPG)). The induction time of hydrate formation was reduced with the presence of cyclopentane (CP). Cyclopentane and methylcyclohexane (MCH) could increase hydrate formation rate, but reduced hydrate storage capacity The higher methylcyclohexane concentration, the lower the hydrate storage capacity. Copyright © 2003 John Wiley & Sons, Ltd.