The aerosol number concentration and size distribution were measured with the newly developed Wide-range Particle Spectrometer in summer and winter of 2006 at the urban site of Jinan City. Here reported the characteri...The aerosol number concentration and size distribution were measured with the newly developed Wide-range Particle Spectrometer in summer and winter of 2006 at the urban site of Jinan City. Here reported the characteristics of fine particles of the different observation seasons. Relative high number concentrations for the particles in the diameter range of 10-500 nm were observed in both seasons. It was found that the dominant number distributed in particle diameter smaller than 100 nm and the percentage over the number concentration of all air particles is much higher than what has been measured in other urban sites over the world. The number mean diameter in summer was much smaller than in winter, strongly suggesting the different origin of ultrafine particles in different seasons. That is, particles in ultrafine mode mainly came from nucleation and new particle formation in summer while from traffic emission in winter. The diurnal variation also supported this point. Number concentration in the diameter range of 10-200 um got their peak values at noontime, well correlated with the mixing ratio of SO2 and the intensity of solar radiation in summer. While in winter, those in the same diameter range showed the main peaks during the traffic hours happened in the morning and evening.展开更多
The main object of the present research is to make a quantitative evaluation on the microstructure of carbon materials in terms of microcrystal. The digitized images acquired from finely pulverized carbon materials un...The main object of the present research is to make a quantitative evaluation on the microstructure of carbon materials in terms of microcrystal. The digitized images acquired from finely pulverized carbon materials under HRTEM at a high magnification were processed by the image processing software so as to extract the fringes of (002) lattice of graphite crystal from the background image, and an FFT-IFFT filtering operation was performed followed by processes as binarization for the image and skeletonization for the fringes. A set of geometrical parameters including position, length and orientation was set up for every lattice fringe by calculating the binarized image. Then, the above obtained fringe parameters were put into an algorithm, which was especially developed for such fringe images so as to find fringes that could be regarded as those belonged to one single graphite microcrystal. The fringe was subjected sequentially to comparing procedures with every other fringe on aspects as parallelism, relative position and spacing, and the above comparisons were repeated till the last fringe. Eventually, the microcrystal size, its stacking number, and the distribution of the microcrystal in the whole sample, as well as other related structure information of such microcrystal in carbon materials were statistically calculated. Such microstructure information at nanometer level may contribute greatly to the interpretation of the properties of carbon materials and a better correlation with the same macrostructure.展开更多
虽然协同过滤可以实现用户的个性化推荐,但是大多数协同过滤及其改进模型未考虑用户和项目等特征,因而不能发掘样本间的非线性关系.与协同过滤相比,深度学习能挖掘丰富的用户兴趣模式,但网络拓扑结构是基于二支决策的方式,忽略了推荐样...虽然协同过滤可以实现用户的个性化推荐,但是大多数协同过滤及其改进模型未考虑用户和项目等特征,因而不能发掘样本间的非线性关系.与协同过滤相比,深度学习能挖掘丰富的用户兴趣模式,但网络拓扑结构是基于二支决策的方式,忽略了推荐样本的难易程度.为了增强模型的非线性表达,同时区分推荐样本的难易,受序贯三支决策的启发,提出序贯三支决策神经网络个性化推荐模型(personalized recommendation model based on sequential three-way decision with single feedforward neural network, STWD-SFNN-PR).首先,为了将高维稀疏特征向量映射为低维稠密的特征向量, STWD-SFNN-PR采用嵌入进行特征处理.其次,在增量式的网络结构中学习推荐样本,使用Adam优化网络参数,并返回难以推荐的样本.再次,利用序贯三支决策增加延迟决策的策略,并在不同的粒度层采用序贯的阈值,从而动态地实现难以推荐样本的划分.最后,为了验证模型的可行性和有效性,选择多种电影推荐数据集进行研究,并选择经典的神经网络推荐、经典的深度学习推荐和最新的三支协同过滤推荐进行对比.实验结果表明, STWD-SFNN-PR具有更优的推荐质量.展开更多
基金Project supported by the National Basic Research Project(973)of China(No.2005CB422203)the National Post-Doctor Foundation of China(No.20060390990).
文摘The aerosol number concentration and size distribution were measured with the newly developed Wide-range Particle Spectrometer in summer and winter of 2006 at the urban site of Jinan City. Here reported the characteristics of fine particles of the different observation seasons. Relative high number concentrations for the particles in the diameter range of 10-500 nm were observed in both seasons. It was found that the dominant number distributed in particle diameter smaller than 100 nm and the percentage over the number concentration of all air particles is much higher than what has been measured in other urban sites over the world. The number mean diameter in summer was much smaller than in winter, strongly suggesting the different origin of ultrafine particles in different seasons. That is, particles in ultrafine mode mainly came from nucleation and new particle formation in summer while from traffic emission in winter. The diurnal variation also supported this point. Number concentration in the diameter range of 10-200 um got their peak values at noontime, well correlated with the mixing ratio of SO2 and the intensity of solar radiation in summer. While in winter, those in the same diameter range showed the main peaks during the traffic hours happened in the morning and evening.
文摘The main object of the present research is to make a quantitative evaluation on the microstructure of carbon materials in terms of microcrystal. The digitized images acquired from finely pulverized carbon materials under HRTEM at a high magnification were processed by the image processing software so as to extract the fringes of (002) lattice of graphite crystal from the background image, and an FFT-IFFT filtering operation was performed followed by processes as binarization for the image and skeletonization for the fringes. A set of geometrical parameters including position, length and orientation was set up for every lattice fringe by calculating the binarized image. Then, the above obtained fringe parameters were put into an algorithm, which was especially developed for such fringe images so as to find fringes that could be regarded as those belonged to one single graphite microcrystal. The fringe was subjected sequentially to comparing procedures with every other fringe on aspects as parallelism, relative position and spacing, and the above comparisons were repeated till the last fringe. Eventually, the microcrystal size, its stacking number, and the distribution of the microcrystal in the whole sample, as well as other related structure information of such microcrystal in carbon materials were statistically calculated. Such microstructure information at nanometer level may contribute greatly to the interpretation of the properties of carbon materials and a better correlation with the same macrostructure.
文摘虽然协同过滤可以实现用户的个性化推荐,但是大多数协同过滤及其改进模型未考虑用户和项目等特征,因而不能发掘样本间的非线性关系.与协同过滤相比,深度学习能挖掘丰富的用户兴趣模式,但网络拓扑结构是基于二支决策的方式,忽略了推荐样本的难易程度.为了增强模型的非线性表达,同时区分推荐样本的难易,受序贯三支决策的启发,提出序贯三支决策神经网络个性化推荐模型(personalized recommendation model based on sequential three-way decision with single feedforward neural network, STWD-SFNN-PR).首先,为了将高维稀疏特征向量映射为低维稠密的特征向量, STWD-SFNN-PR采用嵌入进行特征处理.其次,在增量式的网络结构中学习推荐样本,使用Adam优化网络参数,并返回难以推荐的样本.再次,利用序贯三支决策增加延迟决策的策略,并在不同的粒度层采用序贯的阈值,从而动态地实现难以推荐样本的划分.最后,为了验证模型的可行性和有效性,选择多种电影推荐数据集进行研究,并选择经典的神经网络推荐、经典的深度学习推荐和最新的三支协同过滤推荐进行对比.实验结果表明, STWD-SFNN-PR具有更优的推荐质量.
