分布式稀疏软大间隔聚类

虽然软大间隔聚类(Soft large margin clustering,SLMC)相比其他诸如K-Means等算法具有更优的聚类性能与某种程度的可解释性,然而当面对大规模分布存储数据时,均遭遇了同样的可扩展瓶颈,其涉及的核矩阵计算需要高昂的时间代价。消减此代价的有效策略之一是采用随机Fourier特征变换逼近核函数,而逼近精度所依赖的特征维度常常过高,隐含着可能过拟合的风险。本文将稀疏性嵌入核SLMC,结合交替方向乘子法(Alternating direction method of multipliers,ADMM),给出了一个分布式稀疏软大间隔聚类算法(Distributed sparse SLMC,DS-SLMC)来克服可扩展问题,同时通过稀疏化获得更好的可解释性。

Laboratory Evaluation for Utilization of Phosphogypsum through Carbide Slag Highly-Effective Activating Anhydrous Phosphogypsum

Carbide slag was used as an activator to improve the activity of anhydrous phosphogypsum.Carbide slag could greatly improve the mechanical strength of anhydrous phosphogypsum than K_(2)SO_(4).The compressive strength of 11 wt%carbide slag and 1 wt%K_(2)SO_(4)activated anhydrous phosphogypsum increased greatly to 8.6 MPa at 3 d,and 11.9 MPa at 7 d,and 16.0 MPa at 28 d,respectively.The rate of hydration heat was accelerated and the total hydration heat was increased,and more calcium sulfate dihydrate was formed and cross-linked with other parts which improved the compressive strength of anhydrous phosphogypsum under the effects of different activators.It was indicated that carbide slag was a highly effective and cost-efficient activator.The result provides a highly effective and low-cost method which results in a novel and high value-added method for the utilization of phosphogypsum in the future.

面向本科留学生国际班的“环境科学与工程导论”课程教学改革探索

国际化环境保护专业人才,是我国生态环境保护和生态文明建设的重要力量。建设好《环境科学与工程导论》这一入门启发性课程,对于学生的专业学习和职业发展,至关重要。本文以天津大学为例,从课程团队建设、教学模式等方面,对面向留学生国际班开设的《环境科学与工程导论》课程的教学改革进行了初步探索,以期为环境领域国际化人才培养提供支撑。