The seawater column is typically taken as a homogeneous velocity layer in wide-angle crustal seismic surveys in marine environments. However, heterogeneities in salinity and temperature throughout the seawater layer r...The seawater column is typically taken as a homogeneous velocity layer in wide-angle crustal seismic surveys in marine environments. However, heterogeneities in salinity and temperature throughout the seawater layer result insignificant lateral variations in its seismic velocity, especially in deep marine environments. Failure to compensate for these velocity inhomogeneities will introduce significant artifacts in constructing crustal velocity models using seismic tomography. In this study, we conduct numerical experiments to investigate the impact of heterogeneous seismic velocities in seawater on tomographic inversion for crustal velocity models. Experiments that include lateral variation in seawater velocity demonstrated that the modeled crustal velocities were contaminated by artifacts from tomographic inversions when assuming a homogeneous water layer. To suppress such artifacts, we propose two strategies：（1） simultaneous inversion of water velocities and the crustal velocities;（2） layer-stripping inversion during which to first invert for seawater velocity and then correct the travel times before inverting for crustal velocities. The layer-stripping inversion significantly improves the modeling of variation in seawater velocity when preformed with seismic sensors deployed on the ocean bottom and in the water column. Such strategies improve crustal modeling via wide-angle seismic surveys in deep-marine environment.展开更多
Tin oxide（SnO2） is a promising wide bandgap semiconductor for next generation ultraviolet（UV） nonpolar optoelectronic devices applications.The development of SnO2-based optoelectronic devices is obsessed by its lo...Tin oxide（SnO2） is a promising wide bandgap semiconductor for next generation ultraviolet（UV） nonpolar optoelectronic devices applications.The development of SnO2-based optoelectronic devices is obsessed by its low exciton emission efficiency.In this study,quantum confined SnO2nanocrystals have been fabricated via pulsed laser ablation in water.The SnO2quantum dots（QDs） possess high performance exciton emission at 297-300 nm light in water.The exciton emission intensity and wavelength can be slightly tuned by laser pulse energy and irradiation time.Optical gain has been observed in SnO2QDs.Therefore,SnO2QDs can be a promising luminescence material for the realization of deep UV nanoemitter and lasing devices.展开更多
如何根据不同用户需求将有效信息第一时间呈现给用户是推荐系统的一个重要研究方向。其中,基于深度学习的推荐系统不仅能解决信息冗余还能提升个性化推荐效果,引起了广泛的关注。由于单一的推荐算法往往并不能理想化地满足客户的需求,...如何根据不同用户需求将有效信息第一时间呈现给用户是推荐系统的一个重要研究方向。其中,基于深度学习的推荐系统不仅能解决信息冗余还能提升个性化推荐效果,引起了广泛的关注。由于单一的推荐算法往往并不能理想化地满足客户的需求,本文介绍了一种基于深度学习的组合推荐算法,对基于内容的推荐算法(CB)、矩阵分解推荐算法(MF)和Wide&Deep模型推荐算法进行优化,分别通过岭回归、增加隐式反馈信息和贝叶斯优化的优化方法,并通过转换型的组合方式构成组合推荐算法。实验结果表明,组合算法在Movielens数据集上的推荐效果优于单一模型,评估指标均较于单一模型有基本提升,能够实现精准的内容投放。How to present effective information to users in a timely manner according to their different needs is an important research direction in recommendation systems. Among them, recommendation systems based on deep learning not only solve information redundancy but also improve personalized recommendation performance, which has attracted widespread attention. Due to the fact that a single recommendation algorithm often cannot ideally meet the needs of customers, this article introduces a combination recommendation algorithm based on deep learning, which optimizes content-based recommendation algorithm (CB), matrix factorization recommendation algorithm (MF), and Wide&Deep model recommendation algorithm. The optimization methods include ridge regression, adding implicit feedback information, and Bayesian optimization, and the combination recommendation algorithm is composed of a transformational combination. The experimental results show that the recommendation effect of the combined algorithm on Movielens dataset is better than that of the single model, and the evaluation indicators are basically improved compared with that of the single model, which can achieve accurate content delivery.展开更多
基金supported by the National Natural Science Foundation of China (No.41230318)the Natural Science Foundation of Shandong Province (No.ZR2014DM006)+1 种基金the China Postdoctoral Science Foundation (No.2015M582138)the Scientific Research Foundation for the Returned Overseas Chinese Scholars,Ministry of Education
文摘The seawater column is typically taken as a homogeneous velocity layer in wide-angle crustal seismic surveys in marine environments. However, heterogeneities in salinity and temperature throughout the seawater layer result insignificant lateral variations in its seismic velocity, especially in deep marine environments. Failure to compensate for these velocity inhomogeneities will introduce significant artifacts in constructing crustal velocity models using seismic tomography. In this study, we conduct numerical experiments to investigate the impact of heterogeneous seismic velocities in seawater on tomographic inversion for crustal velocity models. Experiments that include lateral variation in seawater velocity demonstrated that the modeled crustal velocities were contaminated by artifacts from tomographic inversions when assuming a homogeneous water layer. To suppress such artifacts, we propose two strategies：（1） simultaneous inversion of water velocities and the crustal velocities;（2） layer-stripping inversion during which to first invert for seawater velocity and then correct the travel times before inverting for crustal velocities. The layer-stripping inversion significantly improves the modeling of variation in seawater velocity when preformed with seismic sensors deployed on the ocean bottom and in the water column. Such strategies improve crustal modeling via wide-angle seismic surveys in deep-marine environment.
基金the financial support of the project from the National Natural Science Foundation of China(Grant Nos.11004197,11374309,and 11104270)China Postdoctoral Science Foundation Funded Project(Grant No.2013M541847)"Hong Kong Scholars Program"(Grant Nos.XJ2011039,and 201104336)
文摘Tin oxide（SnO2） is a promising wide bandgap semiconductor for next generation ultraviolet（UV） nonpolar optoelectronic devices applications.The development of SnO2-based optoelectronic devices is obsessed by its low exciton emission efficiency.In this study,quantum confined SnO2nanocrystals have been fabricated via pulsed laser ablation in water.The SnO2quantum dots（QDs） possess high performance exciton emission at 297-300 nm light in water.The exciton emission intensity and wavelength can be slightly tuned by laser pulse energy and irradiation time.Optical gain has been observed in SnO2QDs.Therefore,SnO2QDs can be a promising luminescence material for the realization of deep UV nanoemitter and lasing devices.
文摘如何根据不同用户需求将有效信息第一时间呈现给用户是推荐系统的一个重要研究方向。其中,基于深度学习的推荐系统不仅能解决信息冗余还能提升个性化推荐效果,引起了广泛的关注。由于单一的推荐算法往往并不能理想化地满足客户的需求,本文介绍了一种基于深度学习的组合推荐算法,对基于内容的推荐算法(CB)、矩阵分解推荐算法(MF)和Wide&Deep模型推荐算法进行优化,分别通过岭回归、增加隐式反馈信息和贝叶斯优化的优化方法,并通过转换型的组合方式构成组合推荐算法。实验结果表明,组合算法在Movielens数据集上的推荐效果优于单一模型,评估指标均较于单一模型有基本提升,能够实现精准的内容投放。How to present effective information to users in a timely manner according to their different needs is an important research direction in recommendation systems. Among them, recommendation systems based on deep learning not only solve information redundancy but also improve personalized recommendation performance, which has attracted widespread attention. Due to the fact that a single recommendation algorithm often cannot ideally meet the needs of customers, this article introduces a combination recommendation algorithm based on deep learning, which optimizes content-based recommendation algorithm (CB), matrix factorization recommendation algorithm (MF), and Wide&Deep model recommendation algorithm. The optimization methods include ridge regression, adding implicit feedback information, and Bayesian optimization, and the combination recommendation algorithm is composed of a transformational combination. The experimental results show that the recommendation effect of the combined algorithm on Movielens dataset is better than that of the single model, and the evaluation indicators are basically improved compared with that of the single model, which can achieve accurate content delivery.
